Intelligent agent explainable information processing apparatus, method, device, and medium

Abstract

Embodiments of the present application provide an agent explainability information processing device, method, equipment and medium, which can be applied to the field of artificial intelligence technology. The agent explainability information processing device can be applied to the information updating stage, the task planning stage and the task execution stage in the task planning execution process, wherein the device comprises an environment intention updating module, a planning reflection module and an information carrier module. The environment intention updating module is configured to receive environment updating information in the information updating stage of the current round to update planning intention information; the planning reflection module is configured to update planning reflection information according to the updated planning intention information in the task planning stage of the current round; and the information carrier module is configured to record the planning intention information and the planning reflection information in real time in the form of preset record identification data to generate explainability information which is called in the task planning stage of the current round, the task execution stage of the current round and the information updating stage of the next round.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, specifically to the field of image processing technology, and more specifically to an interpretable information processing method, device, equipment, medium and product for an intelligent agent. Background Art

[0002] Artificial Intelligence (AI) is a key driving force behind the new scientific and technological revolution and industrial transformation. It is a new, critical technical discipline that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. As a key component of intelligence science, AI aims to understand the essence of intelligence and produce new intelligent machines (i.e., agents) that can respond in a manner similar to human intelligence.

[0003] Intelligent agents typically need to possess basic characteristics such as subjectivity, purposefulness, recognizability, explainability, value, and feasibility. Explainability refers to the ability of humans to understand and interpret the agent's decision-making process and results. This reflects the transparency of the agent to humans, enabling humans to understand the agent's decision-making process and results, and to trace and explain the agent's decisions and behaviors. However, existing technologies for tracing and explaining the decisions and behaviors of intelligent agents remain difficult to research in this field. The most significant issue is the inability to better explain the agent's real-time information. This leads to inaccurate and untimely information decisions and poor transparency in the interpretability of intelligent agents, making it difficult for humans to understand and interpret the agent's task execution intentions and feedback content. Summary of the Invention

[0004] In view of at least one of the existing problems in the above-mentioned prior art, based on the current development status of explainable artificial intelligence, the embodiments of the present invention aim to provide an explainable information processing device, method, equipment, medium and product for an intelligent body that can realize the explainability of the real-time information of the intelligent body, and provide a multi-stage explainable information recording and calling scheme for the intelligent body based on a loop mechanism. By recording and utilizing the explainable information at each stage in the intelligent body loop program, the information transparency of the intelligent body's decision-making behavior to the developer is achieved, and the language interaction experience between the intelligent body and the user is enhanced, thereby achieving the effect of tracing the source of the intelligent body's behavior and mutual trust between man and machine, realizing the record and utilization of explainable information in the intelligent body planning and decision-making process, and realizing efficient, accurate, real-time and highly transparent intelligent body information explainability.

[0005] One aspect of an embodiment of the present invention provides an interpretable information processing device for an intelligent agent, which is applied to the information update stage, task planning stage and task execution stage in the task planning and execution process, wherein the device includes an environmental intention update module, a planning reflection module and an information carrier module.

[0006] The environmental intention update module is used to receive environmental update information in the information update phase of the current round to update the planning intention information; the planning reflection module is used to update the planning reflection information according to the planning intention information in the task planning phase of the current round; the information carrier module is used to record the planning intention information and planning reflection information in real time in the form of preset record identification data, and generate interpretable information for being called in the task planning phase of the current round, the task execution phase of the current round and the information update phase of the next round.

[0007] According to an embodiment of the present invention, the environment intention update module includes a mental cognition unit. The mental cognition unit is configured to receive environmental perception information of the environment update information in the current round of information update phase to update the mental cognition information.

[0008] According to an embodiment of the present invention, the environment intention update module further includes an intention language update unit. The intention language update unit is configured to receive the environmental sound information of the environment update information in the current round of information update phase to generate a natural language text.

[0009] According to an embodiment of the present invention, the environmental intention update module further includes an intention update unit configured to generate planning intention information based on historical feedback information, mental cognition information, natural language text, and self-value status information during the current round of information update.

[0010] According to one embodiment of the present invention, the planning reflection module includes a batch simulation unit and an asynchronous simulation unit. The batch simulation unit is configured to perform batch simulation based on planning intent information to generate batch simulation information specific to the planning intent information; the asynchronous simulation unit is configured to perform asynchronous simulation based on the planning intent information to generate asynchronous simulation information specific to the planning intent information.

[0011] According to an embodiment of the present invention, the planning reflection module further includes a planning reflection unit configured to update planning reflection information according to the batch simulation information and the asynchronous simulation information.

[0012] According to an embodiment of the present invention, the device further includes an information calling module, which is configured to call interpretable information to perform interactive scenario interpretation and self-intention planning during the task execution phase of the current round.

[0013] Another aspect of an embodiment of the present invention provides an interpretable information processing method for an intelligent agent, which is applied to the information update stage, task planning stage and task execution stage in the task planning execution process, including: receiving environmental update information in the information update stage of the current round to update planning intention information; updating planning reflection information based on the updated planning intention information in the task planning stage of the current round; and recording planning intention information and planning reflection information in real time in the form of preset record identification data to generate interpretable information for being called in the task planning stage of the current round, the task execution stage of the current round and the information update stage of the next round.

[0014] Another aspect of an embodiment of the present invention provides an electronic device comprising one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned method for processing interpretable information of an intelligent agent.

[0015] Another aspect of an embodiment of the present invention provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the above-mentioned interpretable information processing method of the intelligent agent.

[0016] Another aspect of an embodiment of the present invention provides a computer program product, including a computer program, which implements the above-mentioned interpretable information processing method of the intelligent agent when executed by a processor.

[0017] The interpretable information processing device for an intelligent agent provided by an embodiment of the present invention can at least partially solve the problem in related technologies of being unable to better achieve the interpretability of the intelligent agent's real-time information during the execution of intelligent agent task planning, and thus can achieve at least one of the following technical effects:

[0018] (1) By establishing a multi-stage interpretable information recording scheme for intelligent agents based on a loop mechanism, and recording interpretable information at multiple stages in the program loop within the intelligent agent, the transparency of system decision information is achieved, which enables humans to trace the behavior information of the intelligent agent in real time and improves the user's language communication experience.

[0019] (2) Through the record of interpretable information generated based on one’s own value and status, it is possible to realize the record and utilization of interpretable information in the language interaction process based on one’s own value and status.

[0020] (3) The multi-round loop mechanism records historical events and related interpretable information. It can not only explain the information at the current moment, but also integrate historical information to explain the current moment. Therefore, it can enable the intelligent agent to comprehensively utilize the interpretable information of historical conversations and historical events to improve language interaction capabilities.

[0021] In summary, the above-described interpretable information processing device for intelligent agents in embodiments of the present invention can scientifically record interpretability-related information at each stage of the intelligent agent's operation in a real or virtual environment based on a cyclical mechanism, fully leveraging the system's interpretability. Secondly, it enables the intelligent agent to record relevant interpretable information derived from its own value and state based on large-scale model reasoning. Furthermore, during multiple rounds of interaction between the intelligent agent and the environment, historical information can be fully utilized to improve the current interpretable language communication, enhance the intelligent agent's language interaction capabilities, and improve the user's communication experience.

[0022] Therefore, the above-mentioned interpretable information processing device of the intelligent agent in the embodiment of the present invention can support natural language-oriented interpretability. Based on the current development status of interpretable artificial intelligence, it provides an interpretable information recording and information utilization solution in the intelligent agent planning and decision-making process. By recording and utilizing the interpretable information at each stage in the intelligent agent cycle program, the information transparency of the intelligent agent's decision-making behavior to the developer is achieved, and the language interaction experience between the intelligent agent and the user is enhanced, thereby achieving the effect of tracing the source of the intelligent agent's behavior and mutual trust between man and machine.

[0023] It should be understood that the foregoing general description and the following detailed description are merely exemplary and illustrative and are not intended to limit the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The above contents and other objects, features and advantages of the present invention will become more apparent through the following description of the embodiments of the present invention with reference to the accompanying drawings, in which:

[0025] Figure 1 A schematic block diagram of a structure of an interpretable information processing device for an intelligent agent according to an embodiment of the present invention is shown;

[0026] Figure 2 A diagram schematically illustrates an application scenario of an interpretable information processing device for an intelligent agent according to an embodiment of the present invention during a task planning and execution process based on a loop mechanism;

[0027] Figure 3A Schematically illustrates a scene diagram of the interaction between modules of an interpretable information processing device for an intelligent agent according to an embodiment of the present invention during the task planning and execution process based on a loop mechanism;

[0028] Figure 3B A diagram schematically illustrates another application scenario of the interpretable information processing device for an intelligent agent according to an embodiment of the present invention in a task planning and execution process based on a loop mechanism;

[0029] Figure 4A diagram schematically illustrates an application scenario of an interpretable information processing apparatus, method, device, medium, and program product for an intelligent agent according to an embodiment of the present invention;

[0030] Figure 5 A flowchart schematically illustrates a method for processing interpretable information of an intelligent agent according to an embodiment of the present invention; and

[0031] Figure 6 The block diagram of an electronic device suitable for implementing the interpretable information processing method of an intelligent agent according to an embodiment of the present invention is schematically shown.

[0032] The above-mentioned drawings are part of the description of the embodiments of the present invention, illustrating exemplary embodiments of the present invention. Together with the description, the drawings are used to illustrate the principles of the embodiments of the present invention. It should be understood that the above general description of the drawings and the following detailed description are merely exemplary and illustrative and are not intended to limit the scope of the present invention. DETAILED DESCRIPTION

[0033] In order to make the objectives, technical solutions and advantages of the embodiments of the present invention more clearly understood, the spirit of the contents disclosed in the present invention will be clearly illustrated with the accompanying drawings and detailed descriptions below. After understanding the embodiments of the contents of the present invention, any technician in the relevant technical field can change and modify the contents of the present invention based on the techniques taught by the contents of the present invention without departing from the spirit and scope of the contents of the present invention.

[0034] The exemplary embodiments of the present invention and their description are used to explain the present invention, but are not intended to limit the present invention. In addition, elements / components with the same or similar reference numerals used in the drawings and embodiments are used to represent the same or similar parts.

[0035] The terms “first,” “second,” etc. used in the present invention do not particularly refer to an order or sequence, nor are they used to limit the present invention. They are only used to distinguish elements or operations described with the same technical terms.

[0036] The directional terms used in the present invention, such as up, down, left, right, front, or back, are only used with reference to the directions in the accompanying drawings. Therefore, the directional terms used are used to illustrate and not to limit the present invention.

[0037] The terms “include,” “including,” “have,” “contain,” etc. used in the present invention are open-ended terms, meaning including but not limited to.

[0038] The term "and / or" used in the present invention includes any or all combinations of the items mentioned.

[0039] Regarding the present invention, "plurality" includes "two" and "more than two"; regarding the present invention, "plurality of groups" includes "two groups" and "more than two groups".

[0040] The terms "substantially" and "approximately" used in this disclosure are intended to modify any quantity or error that may vary slightly, but such variations or errors do not alter the essence of the quantity. Generally speaking, the range of such variations or errors may be 20% in some embodiments, 10% in some embodiments, 5% in some embodiments, or other values. Those skilled in the art will appreciate that the aforementioned values ​​may be adjusted based on actual needs and are not intended to be limiting.

[0041] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0042] When expressions such as "at least one of A, B, and C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to systems having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, and C, etc.). When expressions such as "at least one of A, B, or C, etc." are used, they should generally be interpreted in accordance with the meaning commonly understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but is not limited to systems having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, and C, etc.). Those skilled in the art should also understand that any transitional conjunctions and / or phrases indicating two or more optional items, whether in the specification, claims, or drawings, should be understood to provide the possibility of including one, either, or both of these items. For example, the phrase "A or B" should be understood to include the possibilities of "A" or "B", or "A and B".

[0043] In existing natural language interaction scenarios, after receiving a QA (question-and-answer) or other command from a user, an agent will respond with a "received" response and execute the command. However, if the command fails, a more general response such as "I can't do that" will be given. Current technology does not proactively inform users of the specific reasons for a network failure, the inability of certain skills to match preconditions, or other errors encountered during task execution. This creates a poor user experience and can feel overly mechanical, failing to reflect the interpretability capabilities of general artificial intelligence.

[0044] In view of at least one of the existing problems in the above-mentioned prior art, based on the current development status of explainable artificial intelligence, the embodiments of the present invention aim to provide an explainable information processing device, method, equipment, medium and product for an intelligent body that can realize the explainability of the real-time information of the intelligent body, and provide a multi-stage explainable information recording and calling scheme for the intelligent body based on a loop mechanism. By recording and utilizing the explainable information at each stage in the intelligent body loop program, the information transparency of the intelligent body's decision-making behavior to the developer is achieved, and the language interaction experience between the intelligent body and the user is enhanced, thereby achieving the effect of tracing the source of the intelligent body's behavior and mutual trust between man and machine, realizing the record and utilization of explainable information in the intelligent body planning and decision-making process, and realizing efficient, accurate, real-time and highly transparent intelligent body information explainability.

[0045] Figure 1 The structure block diagram of the interpretability information processing device 100 of the intelligent agent according to the embodiment of the present invention is schematically shown. Figures 1-3B The interpretability information processing device 100 is further described in detail.

[0046] like Figure 1 As shown, one aspect of an embodiment of the present invention provides an interpretable information processing device 100 for an intelligent agent, which is applied to the information update phase A, the task planning phase B and the task execution phase C during the task planning and execution process. Figure 1 As shown, the interpretability information processing device 100 of the intelligent agent of this embodiment includes an environmental intention updating module 110, a planning reflection module 120 and an information carrier module 130.

[0047] The environment intention update module 110 is used to receive environment update information in the current round of information update phase to update the planning intention information;

[0048] The planning reflection module 120 is used to update the planning reflection information according to the planning intention information during the task planning phase of the current round;

[0049] The information carrier module 130 is used to record planning intention information and planning reflection information in real time in the form of preset record identification data, and generate interpretable information for being called in the current round of task planning stage, the current round of task execution stage and the next round of information update stage.

[0050] In the embodiment of the present invention, Figure 1 The illustrated interpretable information processing device 100 can be widely applied to intelligent agent system architectures based on multi-step loop programs, including intelligent agent systems in simulation environments and physical intelligent robots in real-world environments. In both simulations and real-world environments, the intelligent agent can plan and execute tasks. This can be embodied in a multi-round cycle of "information update - task planning - task execution," enabling the intelligent agent to complete desired actions, such as chatting with a user, fetching water for a user, or cooking.

[0051] like Figure 2 As shown in the diagram of an application scenario of the interpretable information processing device 100, the intelligent agent task planning and execution of an embodiment of the present invention can be understood as the intelligent agent planning and subsequent execution process of the task based on the received and detected environmental interaction information. For example, according to the user's command (the received environmental interaction information), the task is generated (task planning), and the operation is performed according to the generated task (task execution). In this process, the intelligent agent will repeatedly cycle through the above-mentioned task planning and execution process according to the above-mentioned multi-round loop mechanism. Each round of execution will fully realize the three steps of "information update-task planning-task execution", and in the next round of execution, a new round of task planning and execution will be performed based on the previous round of task execution.

[0052] Specifically, if Figure 2 As shown in the figure, based on the above multi-round loop mechanism, the intelligent agent can repeatedly carry out the process of "information update-task planning phase-task execution", which can be divided into three main phases of "information update phase A (UpdateInfostate), task planning phase B (Plan) and task execution phase C (Execute)", which can be implemented in sequence by relying on the intelligent agent's update module, planning module and execution module respectively.

[0053] Specifically, if Figure 2As shown, in the task execution phase C of the previous round, through the agent's operation execution of the execution action on the environment E (Environment), the agent can update the current round of environmental information according to the environmental interaction process and enter the information update phase A; in the information update phase A of the current round, the planning intention information of the agent is generated through the updated environmental information (that is, the execution intention Intent of the next round), and the task planning phase B is entered; in the task planning phase B of the current round, the execution action is planned through the agent execution intention of the above planning intention information, and the optimal execution trajectory (traj_list) is generated, and the task execution phase C is entered; in the task execution phase C of the current round, the corresponding operation action (Action) is performed on the environmental target according to the above optimal execution trajectory, thereby completing the entire process of task planning and execution of the current round.

[0054] Furthermore, the update module (Update infostate) corresponds to the information update stage A, and is mainly responsible for reasoning based on environmental cognitive information (belief_pg), which is used to complete the external environmental interaction information such as vision, language, and sound in the interpretable information processing device 100 (infostate) and the corresponding cognitive reasoning information. For example, it can include the agent's inference information on unobserved information, inference information on the future state of things, and the mental state information of the agent based on the inference.

[0055] The planning module (Plan) corresponds to the task planning phase B. Using a planning framework based on the UV architecture (Skill U - Value), the planning module receives input information including the environment's cognitive state (belief state), including the agent's current cognition and focus. It also includes a U library describing the agent's skill repertoire (e.g., Rubik's Cube strategies and the possible outcomes of different strategies). Furthermore, it includes an action library (Action library), which is called upon by different U libraries and forms part of the U library strategy. Finally, it includes a V library (V Value library), which represents the agent's value evaluation. The planning module selects the skill library U and its related parameters, ultimately outputting a decision (i.e., the execution path for skill library U, such as traj_list). Therefore, during the current round of task planning, the planning module's decisions can be based on the different skills U selected after simulation.

[0056] The execution module (execute) corresponds to the task execution phase C, and is mainly responsible for the execution of the above-mentioned planning decisions. During its execution, it can realize the real-time update of relevant environmental cognitive information (corresponding to the next round of information update phase). At the same time, during the execution process, it can simultaneously perform sub-tasks such as learning, reflection, and decision adjustment based on the latest information update content. That is, the execution module can control the output of the intelligent agent's task-related actions for the environment while updating its own state and model.

[0057] like Figures 1-3A As shown, the environmental intention update module 110 can be equivalent to a component of the aforementioned update module (update). In the information update phase A, the environmental update information can be the interaction information generated by the current round of interaction between the agent and the environment, such as environmental visual information (such as indoor people, space, objects, size, shape, color, etc.), environmental sound information (such as voiceprints, speech, noise, etc.), and environmental cognitive information (such as distinguishing who a person is, what a person is doing, and what a person may do, etc.) detected by the agent through various sensors through interactive actions such as "seeing", "hearing", and "touching". In addition, the environmental update information can also include environmental interaction information passively received by the agent, such as command information input by the user through the agent's interactive display touch interface.

[0058] Planning intention information can be the execution intention corresponding to the current round of information update phase, generated by the agent through reasoning and cognition based on the aforementioned environmental update information. This execution intention can determine the agent's task goal in the current round of task execution phase. Because relevant historical planning intention information already exists in the previous round of information update phase, the current round of information update phase A is a further update of this historical planning intention information, such as adding a new execution intention and abandoning the previous execution intention based on this, or modifying the previous execution intention to form a new execution intention.

[0059] like Figures 1-3A As shown, the planning reflection module 120 can be understood as a component of the aforementioned planning module (plan). In the task planning phase B, the planning reflection information can generally be the execution reflection information generated by the intelligent agent through simulation execution according to the execution intention determined by the planning intention information. Specifically, it is to call the execution operator to perform execution simulation on each execution path (or execution trajectory) formed in a virtual simulation environment, and select the optimal execution path based on the evaluation (value) of the simulation results generated by the execution simulation. Among them, the process of executing the simulation and performing the path evaluation can be understood as "reflection" based on the execution intention, and the output optimal execution path and related content can be understood as the main components of the planning reflection information.

[0060] like Figures 1-3AAs shown, the information carrier module 130 (context) can be equivalent to the carrier for information storage and call management of each module unit such as the above-mentioned environmental intention update module 110 and the planning reflection module 120. It is the main information carrier of the above-mentioned intelligent body's cycle stage and can be used to record almost all effective information of each link, each stage and external environment feedback of the intelligent body, including the interpretability information mentioned in the embodiment of the present invention.

[0061] The preset record identification data format can be the specified data format (such as record identification) in which the planning intention information and planning reflection information are recorded in the information carrier module 130. It can usually have a certain record identification to record different types of information separately, so as to realize the classified recording and management (such as calling) of the agent information at different stages. Taking the currently applied agent framework process as an example, Figure 3A The illustrated information carrier module 130 (context) can be the primary information management module within the interpretable information processing device 100 (infostate), responsible for recording and cyclically transmitting historical information. The different interpretable information within this information carrier module 130 can be various types of information stored by the agent at different stages of the loop, in the form of data with preset record identifiers similar to the record identifier. The interpretable information generated by the agent at each stage of the loop system can primarily fall into three categories: improper operator logic (e.g., lfp_warm up), non-compliance with action preconditions (e.g., error codes), and inconsistent value orientation (e.g., low evaluations resulting in value conflicts leading to the abandonment of corresponding actions). The recording and retrieval of interpretable information can occur during the entire loop process, including the information update phase A, the task planning phase B, and the task execution phase.

[0062] In summary, in the interpretable information processing device 100 of the intelligent body in the embodiment of the present invention, by constructing a multi-stage interpretable information recording scheme for the intelligent body based on a loop mechanism, the information carrier module 130 can be used to realize multi-stage interpretable information recording and calling in the internal program loop of the intelligent body, thereby ensuring that the system decision information is more transparent, allowing humans to trace the behavior information of the intelligent body, thereby significantly improving the user's language communication experience.

[0063] Moreover, based on the multi-round cycle mechanism to record historical events and related explainable information, it can not only explain the information at the current moment, but also explain the present moment by integrating historical information, thereby ensuring the real-time and accuracy of explainable information at the same time, and achieving a more transparent intelligent agent explainability capability in a truly sense, thereby achieving the effect of tracing the source of intelligent agent behavior and mutual trust between humans and machines.

[0064] like Figures 1-3BAs shown, according to one embodiment of the present invention, the environmental intention updating module 110 includes a mental cognition unit 311 .

[0065] The mental cognition unit 311 is configured to receive environmental perception information of the environment update information in the current round of information update phase A to update the mental cognition information.

[0066] like Figure 3A As shown, the mental cognition unit 311 (believe step) corresponding to the information update stage A can process the environmental perception information such as visual perception and auditory perception in the environmental update information received by the intelligent agent from the external environment. The environmental perception information can also include relevant information on whether the task goals were achieved normally in historical behaviors (such as the previous round of action execution process), and then update its own mental cognition information according to the relevant environmental update information (such as env_info_update) received from the outside world.

[0067] Among them, the mental cognitive information may include the external environment interaction information in the environmental update information and the corresponding cognitive reasoning information, for example, it may include the agent's inference information on unobserved information, the inference information on the future state of things, and the mental state information of the agent based on the inference update.

[0068] In the mental cognition unit 311 corresponding to this information update phase, all external feedback on the agent's previous decision-making behavior can be recorded. For example, the generated error code can be recorded as mental cognition information in the information carrier module 130 based on the preset record identification data format of evt_record. Specifically, for example, when the agent attempts the target task of unscrewing a bottle cap, it finds that the bottle cap is too tight to be unscrewed during the execution of the action. The information generated during the execution process is then transmitted to the agent as environmental update information, and its environmental perception information (such as the bottle cap is too tight to be unscrewed) is recorded as mental cognition information in the information carrier module 130 (Context).

[0069] like Figures 1-3B As shown, according to an embodiment of the present invention, the environment intention updating module 110 further includes an intention language updating unit 312 .

[0070] The intended language updating unit 312 is configured to receive the ambient sound information of the environment update information in the current round of information updating phase A to generate a natural language text.

[0071] like Figure 3AAs shown, the mind language update unit 312 (mind_nlp_update) corresponding to information update phase A can extract environmental update information from the agent's interaction with the external environment E to generate environmental sound information. This environmental sound information can be environmental sound data (such as conversational speech, voiceprint, etc.) during the agent-environment interaction. The mind language update unit 312 can also selectively convert this environmental sound information to generate natural language text, such as the text of a conversation between a human and an agent. This natural language text can be generated through natural language processing (NLP) to more accurately reflect human language expression (current NLP message). This natural language text information can be recorded as natural language text information in the information carrier module 130 based on the preset record identification data format of nlp_record. Therefore, this natural language text information can be used to verify and track interactive information of language communication between agents and humans, and between agents.

[0072] like Figures 1-3B As shown, according to an embodiment of the present invention, the environment intention updating module 110 further includes an intention updating unit 313 .

[0073] The intention updating unit 313 is used to generate planning intention information based on historical feedback information, mental cognition information, natural language text and self-value status information in the current round of information updating phase A.

[0074] like Figure 3AAs shown, the intent update unit 313 (intent update) corresponding to the information update stage A is the last processing unit corresponding to this stage. The intent update unit 313 can integrate all the previously available information to perform comprehensive information reasoning and judgment to generate planning intention information corresponding to the execution intention. Among them, these existing information may include the execution feedback information (i.e., historical feedback information) generated by the execution of historical target tasks in the previous round of the cycle, and may also include the environmental update information of the aforementioned external feedback, and may also include the mental cognition information updated by the mental cognition unit 311, the natural language text updated by the intention language update unit 312, and the self-value status information. Among them, the self-value status information may include the value information generated by the evaluation of the agent's own behavior and the status information related to the agent's own behavior status in the current round, etc., which can be mainly used to trigger new intentions (intent). For example, when an agent is performing a target task of moving from the living room to the bedroom to help a user pick up an item, it encounters a stool in the aisle, preventing it from completing the target task according to the original action. At this point, the agent can replan the task based on its own state (such as the distance from the obstacle) and its own behavior evaluation (such as whether its current behavior prevents it from passing smoothly), and obtain a new target subtask for the current round of bypassing or removing the obstacle. This corresponds to the process of generating an execution intention. The planning intention information generated by the above historical feedback information, mental cognition information, natural language text, and self-value state information can be recorded in the information carrier module 130 in the form of preset record identification data of int_record and evt_record, respectively.

[0075] Therefore, the aforementioned planning intention information can include two interpretable types of information: intent and value. Intent represents the agent's inner thoughts and motivations. For example, when receiving external conversation information and having a high level of friendship with the speaker, the agent may intend to speak. Value represents the reason for not executing the target task when a situation conflicts with the agent's own values ​​or status. For example, if the agent is ordered to destroy the fish tank, this action may conflict with its existing values ​​of protecting user property or even caring for animals, so the agent may choose not to execute the task. These two types of records at this stage explain the agent's thoughts and inner thoughts in specific situations, making it easier for humans to observe the agent's inner thoughts.

[0076] Therefore, the interpretable information record of the planning intention information generated based on its own value and state can realize the recording and utilization of interpretable information based on the language interaction process of its own value and state, so that. Moreover, in this process, historical events and other related interpretable information can be recorded based on a multi-round cycle mechanism, which can not only explain the information at the current moment, but also explain the current moment by integrating historical information, giving full play to the system interpretability and realizing efficient, accurate, and real-time highly transparent intelligent agent information interpretability.

[0077] like Figures 1-3B As shown, according to an embodiment of the present invention, the planning reflection module 120 includes a batch simulation unit 321 and an asynchronous simulation unit 322 .

[0078] The batch simulation unit 321 is used to perform batch simulation according to the planning intention information to generate batch simulation information for the planning intention information;

[0079] The asynchronous simulation unit 322 is used to perform asynchronous simulation according to the planning intention information to generate asynchronous simulation information for the planning intention information.

[0080] like Figure 3A As shown, the batch simulation unit 321 (batch simulate) can perform batch planning simulation for the above-mentioned planning intention information in the current round of loop, and can simultaneously perform batch simulation processing of multiple planning execution paths corresponding to the above-mentioned planning intention information. Specifically, for each execution path, the corresponding execution operator can be called, and simulation execution can be performed in a simulation environment to obtain corresponding simulation results, that is, batch simulation information.

[0081] Unlike the aforementioned batch simulation, asynchronous simulation unit 322 (asynchronous simulation unit 322) can perform independent asynchronous planning simulations based on the aforementioned planning intent information. This asynchronous simulation process is not constrained by the aforementioned agent loop process and is independent of it, without creating any time conflicts with it. During this asynchronous simulation process, each execution path can be simulated in the simulation environment by calling the corresponding execution operator to obtain the corresponding simulation results, i.e., asynchronous simulation information.

[0082] During the corresponding task planning phase B, batch simulation and asynchronous simulation can start simultaneously. However, while the batch simulation completes the current round of simulation, the asynchronous simulation may still be simulating the previous round. This means that the asynchronous simulation is not restricted by the current round of task planning phase B and can be independent of it. When a batch simulation completes, the asynchronous simulation may also complete at the same time, generating both batch and asynchronous simulation information.

[0083] Specifically, when a user assigns the agent a "water collection" target task, the intent update unit can generate the corresponding planning intent information (Intent) for this task. During the task planning phase, the batch simulation unit 321 can plan action execution paths based on this planning intent information. It can simultaneously generate and plan multiple action execution paths (e.g., a complete execution path of "find a water dispenser and a cup - walk toward the cup - pick up the cup - turn around - walk toward the water dispenser - place the cup in the water inlet - turn on the water inlet - wait for the cup to be full - turn off the water inlet - pick up the cup - turn around - walk toward the user") and perform batch simulations on these execution paths to obtain batch simulation information corresponding to that phase. For specific, non-timely execution intents related to planning intent information, the asynchronous simulation unit 322 does not need to consider the execution simulation phase of the batch simulation unit 321. This is completely independent of the agent's "update-plan-execute" loop mechanism, allowing for parallel asynchronous planning and path simulation. For example, in the above-mentioned "getting water" task, after the agent "walks to the water dispenser - puts the cup to the water inlet - turns on the water inlet switch", it finds that the water dispenser is not plugged in. At this time, the agent will generate a new "plug in the water dispenser" task. At this time, this "plug in the water dispenser" task can also correspond to a new execution path, such as "find the water dispenser wiring plug and the nearest socket - turn around - walk to the plug - take the plug - turn around - walk to the socket - take the plug and insert it into the socket". At this time, the path simulation of the new "plug in the water dispenser" task can be understood as an asynchronous simulation relative to the above-mentioned "getting water" task, and then obtain the corresponding asynchronous simulation information.

[0084] like Figures 1-3B As shown, according to an embodiment of the present invention, the planning reflection module 120 further includes a planning reflection unit 323 .

[0085] The planning reflection unit 323 is configured to update the planning reflection information according to the batch simulation information and the asynchronous simulation information.

[0086] like Figure 3A As shown, corresponding to the task planning phase B, the planning reflection unit 323 (reflect) can perform information filtering and reflection after completing the execution simulation in the planning phase (plan) according to the batch simulation information and the asynchronous simulation information.

[0087] Among them, since in the aforementioned batch simulation and asynchronous simulation processes, multiple batch simulation and asynchronous simulation action execution paths can be obtained as decision trajectories (trajectory) of the subsequent execution stage (execution), each trajectory can include a subsequent series of execution actions and action parameters. These trajectories will be evaluated or assessed (Value) through the planning reflection unit 323. Specifically, the simulation execution results of each trajectory can be evaluated separately. For example, it can be evaluated based on the expected value of energy consumption of the trajectory execution, or it can be comprehensively scored based on other preset standards to form an evaluation result.

[0088] Based on the evaluation results, the planning and reflection unit 323 can also perform pruning and filtering of decision trajectories (removing trajectories with the lowest assessed value or those with value conflicts), ultimately selecting the optimal decision trajectory for execution based on the scoring ranking. Specifically, among multiple simulated decision trajectories, unreasonable trajectories are eliminated due to their low scores. Simultaneously, interpretability-related information (such as lfp_warm up, Plan, etc.) is extracted from these simulated unreasonable and reasonable trajectories and recorded in the form of preset record identifier data (such as evt_record) via the information carrier module 130, greatly improving the interpretability of the agent's "non-execution" or "execution" of certain decisions. For example, if a user commands the agent to perform the task of "turning on the TV," and the TV in the environment is already turned on, the simulated trajectory score associated with turning on the TV will be low. Ultimately, the agent may choose not to perform any tasks or to perform other tasks based on its own value state. In this case, the command to turn on the TV is not executed because the TV is already on. This reason and the execution operation can be recorded as interpretable information in the information carrier module 130.

[0089] Therefore, through the multi-stage explainable information recording scheme in the internal program loop of the intelligent agent, the transparency of system decision-making information can be effectively achieved, allowing humans to trace the behavior information of the intelligent agent, greatly improving the user's language communication experience.

[0090] Among them, the interpretability function of the interpretability information processing device can be widely used in Figure 2 Under the framework of the intelligent body loop processing system shown, the recording and utilization of interpretable information can be realized as a whole.

[0091] By this means, the above-mentioned interpretable information processing device of the embodiment of the present invention can provide a technology that enables an intelligent agent to comprehensively utilize the interpretable information of historical conversations and historical events to improve language interaction capabilities, and adopts a multi-round loop mechanism to record historical events and related interpretable information, which can not only explain the information at the current moment, but also effectively integrate historical information to explain the present moment.

[0092] like Figures 1-3B As shown, according to one embodiment of the present invention, the interpretability information processing device 100 further includes an information calling module 140.

[0093] The information calling module 140 is used to call the interpretability information to perform interactive scenario interpretation and self-intention planning in the task execution phase C of the current round.

[0094] like Figure 3B As shown, in an embodiment of the present invention, the interpretable information recorded by an intelligent agent based on a cyclic mechanism architecture during each stage of "information update-task planning-task execution" can be utilized in two aspects: the first is for contextual interpretation during verbal communication between the intelligent agent and humans (i.e., external interpretation, which can be understood as interactive contextual interpretation); the second is for reference to historical information when the intelligent agent infers its own intentions and generates planning intention information (i.e., internal intention generation, which can be understood as self-intention planning). The former interactive contextual interpretation is aimed at the external environment E and can explain the reasons for the intelligent agent's decisions; the latter self-intention planning is internal and can assist the intelligent agent in analyzing historical processes and generating new intentions.

[0095] The agent-generated intentions simulate the inner drive of humans after receiving external feedback. The innovative feature of this architecture is that the agent can process related tasks using its own value state information. Due to the cyclical mechanism of the agent system architecture, the information carrier module 130 (context) can record interpretable information at each stage of multiple cycles. This allows the agent to combine its own value and state during information update phase A (e.g., intent_update) and use various types of information such as historical behavior, dialogue, and actions as input to a preset language model (e.g., Large Language Model, LLM) to generate different intentions. This system architecture based on the multi-cycle recording of interpretable information by the information carrier module 130 ensures the integrity of historical information while also ensuring the rationality and comprehensiveness of the agent's interpretable information, thus playing a key role in generating reasonable agent intentions.

[0096] For instructions given by users to an agent, the agent can choose whether to execute them based on its own value and status. If the agent chooses not to execute an instruction, the decision-making system can use the collected interpretable information through the information call module 140 to provide an explanation. This system architecture, which records interpretable information in multiple cycles, integrates historical information to present the agent's true inner monologue. Using a preset language model, it provides a final, summary response that aligns with human expression habits. This allows users to more clearly demonstrate the true reasons behind the agent's decision-making, enhancing the language interaction experience.

[0097] Therefore, whether in real or virtual environments, the loop-based intelligent agent can fully utilize the system's interpretability capabilities by scientifically recording interpretability-related information at each stage of its operation. Furthermore, the intelligent agent can infer relevant interpretability information based on its own value and state based on a preset language model. This allows the intelligent agent to fully utilize historical information in multiple rounds of interaction with the environment, improving the current interpretable language communication process, significantly enhancing the intelligent agent's language interaction capabilities, and improving the user's communication experience.

[0098] In summary, the interpretable information processing device for an intelligent agent provided by the embodiments of the present invention achieves at least one of the following technical effects:

[0099] (1) By establishing a multi-stage interpretable information recording scheme for intelligent agents based on a loop mechanism, and recording interpretable information at multiple stages in the program loop within the intelligent agent, the transparency of system decision information is achieved, which enables humans to trace the behavior information of the intelligent agent in real time and improves the user's language communication experience.

[0100] (2) Through the record of interpretable information generated based on one’s own value and status, it is possible to realize the record and utilization of interpretable information in the language interaction process based on one’s own value and status.

[0101] (3) The multi-round loop mechanism records historical events and related interpretable information. It can not only explain the information at the current moment, but also integrate historical information to explain the current moment. Therefore, it can enable the intelligent agent to comprehensively utilize the interpretable information of historical conversations and historical events to improve language interaction capabilities.

[0102] In summary, the above-described interpretable information processing device for intelligent agents in embodiments of the present invention can scientifically record interpretability-related information at each stage of the intelligent agent's operation in a real or virtual environment based on a cyclical mechanism, fully leveraging the system's interpretability. Secondly, it enables the intelligent agent to record relevant interpretable information derived from its own value and state based on large-scale model reasoning. Furthermore, during multiple rounds of interaction between the intelligent agent and the environment, historical information can be fully utilized to improve the current interpretable language communication, enhance the intelligent agent's language interaction capabilities, and improve the user's communication experience.

[0103] Therefore, the above-mentioned interpretable information processing device of the intelligent agent in the embodiment of the present invention can support natural language-oriented interpretability. Based on the current development status of interpretable artificial intelligence, it provides an interpretable information recording and information utilization solution in the intelligent agent planning and decision-making process. By recording and utilizing the interpretable information at each stage in the intelligent agent cycle program, the information transparency of the intelligent agent's decision-making behavior to the developer is achieved, and the language interaction experience between the intelligent agent and the user is enhanced, thereby achieving the effect of tracing the source of the intelligent agent's behavior and mutual trust between man and machine.

[0104] According to an embodiment of the present invention, any multiple modules among the environmental intention update module 110, the planning reflection module 120, the information carrier module 130, and the information call module 140 can be combined into a single module for implementation, or any one of these modules can be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules can be combined with at least part of the functionality of other modules and implemented in a single module. According to an embodiment of the present invention, at least one of the environmental intention update module 110, the planning reflection module 120, the information carrier module 130, and the information call module 140 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application-specific integrated circuit (ASIC), or can be implemented in hardware or firmware by any other reasonable method of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in an appropriate combination of any of them. Alternatively, at least one of the environmental intention updating module 110 , the planning reflection module 120 , the information carrier module 130 and the information calling module 140 may be at least partially implemented as a computer program module, which may perform corresponding functions when executed.

[0105] Based on the above-mentioned interpretable information processing device of an intelligent agent, the present invention also provides an interpretable information processing method of an intelligent agent. Figure 4-Figure 5 The method is described in detail.

[0106] One aspect of an embodiment of the present invention provides an interpretable information processing method for an intelligent agent, which is applied to the information update stage, task planning stage and task execution stage in the task planning execution process, wherein the method includes: receiving environmental update information in the information update stage of the current round to update planning intention information; updating planning reflection information based on the updated planning intention information in the task planning stage of the current round; and recording planning intention information and planning reflection information in real time in the form of preset record identification data to generate interpretable information for being called in the task planning stage of the current round, the task execution stage of the current round and the information update stage of the next round.

[0107] Figure 4 The application scenario diagram of the interpretable information processing device, method, equipment, medium and program product of the intelligent agent according to the embodiment of the present invention is schematically shown.

[0108] like Figure 4As shown, an application scenario 400 according to this embodiment may include terminal devices 401, 402, 403, a network 404, and a server 405. The network 404 is used as a medium for providing a communication link between the terminal devices 401, 402, 403 and the server 405. The network 404 may include various connection types, such as wired or wireless communication links or optical fiber cables.

[0109] Users can use terminal devices 401, 402, and 403 to interact with server 405 via network 404 to receive or send messages, etc. Terminal devices 401, 402, and 403 can be installed with various communication client applications, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0110] The terminal devices 401 , 402 , and 403 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0111] Server 405 may be a server that provides various services, such as a backend management server (for example only) that supports websites browsed by users using terminal devices 401, 402, and 403. The backend management server may analyze and process received data such as user requests, and provide feedback (e.g., web pages, information, or data obtained or generated based on user requests) to the terminal device.

[0112] It should be noted that the interpretability information processing method of the intelligent agent provided in the embodiment of the present invention can generally be executed by the server 405. Accordingly, the interpretability information processing device of the intelligent agent provided in the embodiment of the present invention can generally be set in the server 405. The interpretability information processing method of the intelligent agent provided in the embodiment of the present invention can also be executed by a server or server cluster that is different from the server 405 and can communicate with the terminal devices 401, 402, 403 and / or the server 405. Accordingly, the interpretability information processing device of the intelligent agent provided in the embodiment of the present invention can also be set in a server or server cluster that is different from the server 405 and can communicate with the terminal devices 401, 402, 403 and / or the server 405.

[0113] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0114] The following will be based on Figure 4 The scene described by Figure 5The interpretable information processing method of the intelligent agent of the disclosed embodiment is described in detail.

[0115] like Figure 5 As shown, one aspect of an embodiment of the present invention provides an interpretable information processing method for an intelligent agent, which is applied to the information updating stage, task planning stage and task execution stage in the task planning execution process, including operations S501 to S503.

[0116] In operation S501, the environment update information is received in the current round of information update phase to update the planning intention information. In one embodiment, the environment intention update module 110 described above can be used to implement the execution, which will not be repeated here.

[0117] In operation S502, in the current round of task planning, the planning reflection information is updated according to the updated planning intention information. In one embodiment, this can be implemented by the planning reflection module 120 described above, which will not be described in detail here.

[0118] In operation S503, planning intention information and planning reflection information are recorded in real time using a preset recording identification data format to generate interpretable information for use during the current round of task planning, the current round of task execution, and the next round of information update. In one embodiment, this can be implemented using the information carrier module 130 described above, which will not be further described here.

[0119] The intelligent agent can be the executor of the above-mentioned interpretable information processing method of the embodiment of the present invention, or it can be an executor controlled by the interpretable information processing method. Specifically, it can be a humanoid intelligent robot or other AI device, which usually has its own actuator to complete specific action tasks. For example, a humanoid robot can use a mechanical manipulator to complete the action task of picking up an object, or even perform action tasks such as packaging products on an industrial production line. Among them, these action tasks can be used as the target execution tasks of the intelligent agent. In order to complete these target execution tasks, the intelligent agent needs to complete a series of execution actions in sequence.

[0120] The agent can monitor its environment and state in real time, confirming the progress of its current target task. Specifically, it can assess its spatial environment and state during each iteration. The current execution state can be at least one of the state information of the agent's currently detected action and the state of its environment. The state information of the action can include the name of the currently executed action and execution information corresponding to the action (such as wrist torque, rotation angle, expected execution time, and the spatial location of each key point of the finger). The state of the environment can include information about people and objects in the space surrounding the agent at the current moment, such as the size, location (distance), and even shape and color of obstacles on the agent's route. It can also include dynamic information about people in the surrounding environment, such as their movement speed and facial expression. Compared to the historical environmental information obtained from the previous environmental detection, the environmental information obtained from the current environmental detection can be used as the aforementioned environmental update information for related operations in the information update phase.

[0121] Figure 6 The block diagram of an electronic device suitable for implementing the interpretable information processing method of an intelligent agent according to an embodiment of the present invention is schematically shown.

[0122] The above-mentioned electronic device provided by an embodiment of the present invention includes one or more processors and a memory, and the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned interpretable information processing method of the intelligent agent.

[0123] like Figure 6 As shown, an electronic device 600 according to an embodiment of the present invention includes a processor 601, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 602 or programs loaded from a storage unit 608 into a random access memory (RAM) 603. The processor 601 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or related chipsets and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 601 may also include onboard memory for caching purposes. The processor 601 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0124] Various programs and data required for the operation of the electronic device 600 are stored in the RAM 603. The processor 601, ROM 602, and RAM 603 are connected to each other via a bus 604. The processor 601 executes the programs in the ROM 602 and / or RAM 603 to perform various operations according to the method flow of the embodiment of the present invention. It should be noted that the programs may also be stored in one or more memories other than the ROM 602 and RAM 603. The processor 601 may also execute the programs stored in the one or more memories to perform various operations according to the method flow of the embodiment of the present invention.

[0125] According to an embodiment of the present invention, electronic device 600 may further include an input / output (I / O) interface 605, which is also connected to bus 604. Electronic device 600 may also include one or more of the following components connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including devices such as a cathode ray tube (CRT), liquid crystal display (LCD), and speakers; a storage section 608 including a hard disk; and a communication section 609 including a network interface card such as a LAN card or modem. Communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. Removable media 611, such as a magnetic disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed in drive 610 as needed, so that computer programs read from the removable media can be installed into storage section 608 as needed.

[0126] The present invention also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to execute the above-mentioned interpretable information processing method of the intelligent agent.

[0127] The computer-readable storage medium may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, and when executed, implements the method according to the embodiment of the present invention.

[0128] According to an embodiment of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present invention, a computer-readable storage medium may include the ROM 602 and / or RAM 603 described above, and / or one or more memories other than ROM 602 and RAM 603.

[0129] An embodiment of the present invention also includes a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the above-mentioned interpretable information processing method of the intelligent agent.

[0130] The computer program includes program codes for executing the method shown in the flowchart. When the computer program product is run in a computer system, the program codes are used to enable the computer system to implement the method provided by the embodiment of the present invention.

[0131] The computer program executes the above functions defined in the system / device of the embodiment of the present invention when the computer program is executed by the processor 601. According to the embodiment of the present invention, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0132] In one embodiment, the computer program may be stored on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may be transmitted and distributed in the form of a signal on a network medium, downloaded and installed via the communication portion 609, and / or installed from a removable medium 611. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to wireless, wired, or any suitable combination thereof.

[0133] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 609 and / or installed from a removable medium 611. When the computer program is executed by the processor 601, the above-described functions defined in the system of the embodiment of the present invention are performed. According to the embodiment of the present invention, the systems, devices, means, modules, units, etc. described above can be implemented by computer program modules.

[0134] According to an embodiment of the present invention, the program code for executing the computer program provided by the embodiment of the present invention can be written in any combination of one or more programming languages. Specifically, these computer programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C" or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect via the Internet).

[0135] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0136] In addition, all actions of acquiring information, signals or data in the present invention are carried out in compliance with the relevant data protection laws, regulations and policies of the country where they are located, and with the authorization given by the owner of the corresponding device.

[0137] Those skilled in the art will appreciate that various combinations and / or combinations of features described in the various embodiments and / or claims of the present invention may be made, even if such combinations and / or combinations are not explicitly described in the present invention. In particular, various combinations and / or combinations of features described in the various embodiments and / or claims of the present invention may be made, without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0138] The above describes embodiments of the present invention. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of the present invention. Although each embodiment has been described separately above, this does not mean that the measures in each embodiment cannot be advantageously used in combination. The scope of the present invention is defined by the appended claims and their equivalents. Without departing from the scope of the present invention, those skilled in the art may make various substitutions and modifications, which are intended to fall within the scope of the present invention.

Claims

1. An interpretable information processing device for an intelligent agent, applied to the repeated cycle of information updating, task planning, and task execution in the task planning and execution process of an intelligent agent in a natural language interaction scenario, characterized in that: include: An environmental intention update module is configured to receive environmental update information during the information update phase of the current round to update planning intention information, wherein the environmental intention update module includes a mental cognition unit, an intentional language update unit, and an intention update unit, wherein the intention update unit is configured to generate planning intention information during the information update phase of the current round based on historical feedback information, mental cognition information updated by the mental cognition unit, natural language text generated by the intentional language update unit, and self-value status information; wherein the planning intention information includes intention information and value information, wherein intention information is the thoughts and motivations in the mind of the intelligent body, and value information is the reason for choosing not to perform the target task when a situation conflicts with its own value or status; a planning reflection module, configured to update planning reflection information according to the planning intention information during the task planning phase of the current round; and An information carrier module is configured to record the planning intention information and planning reflection information in real time in a preset record identification data format, generating interpretable information for use during the task planning phase of the current round, the task execution phase of the current round, and the information update phase of the next round. The preset record identification data format is a designated data format for recording the planning intention information and planning reflection information in the information carrier module, and has record identifications for separately recording different types of information. The interpretable information includes three categories: inappropriate operator logic, non-compliance with action preconditions, and non-compliance with value orientation. The information calling module is used to call the interpretable information for interactive scenario interpretation and self-intention planning during the task execution phase of the current round. Interactive scenario interpretation is the scenario interpretation during language communication between the intelligent agent and humans; self-intention planning is the reference to historical information when the intelligent agent makes self-intention inference and generates planning intention information.

2. The device according to claim 1, characterized in that The mental cognition unit is configured to receive the environmental perception information of the environmental update information in the information update phase of the current round to update the mental cognition information.

3. The device according to claim 2, characterized in that The intended language updating unit is configured to receive the environmental sound information of the environmental update information during the information updating phase of the current round to generate a natural language text.

4. The device according to claim 1, characterized in that The planning reflection module includes: a batch simulation unit, configured to perform batch simulation according to the planning intention information to generate batch simulation information for the planning intention information; An asynchronous simulation unit is used to perform asynchronous simulation according to the planning intention information to generate asynchronous simulation information for the planning intention information.

5. The device according to claim 4, characterized in that The planning reflection module also includes: The planning reflection unit is used to update the planning reflection information according to the batch simulation information and the asynchronous simulation information.

6. A processing method for an interpretable information processing device of an intelligent agent according to any one of claims 1 to 5, applied to the information update stage, task planning stage and task execution stage in the task planning and execution process, characterized in that: include: receiving environment update information in the information update phase of the current round to update planning intention information; In the task planning phase of the current round, updating the planning reflection information according to the updated planning intention information; as well as The planning intention information and planning reflection information are recorded in real time in the form of preset record identification data to generate interpretable information that is called in the task planning stage of the current round, the task execution stage of the current round, and the information update stage of the next round.

7. An electronic device comprising: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors are caused to perform the method of claim 6.

8. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, causes the processor to perform the method according to claim 6.

Images