WO2025199736A1 - 机器人的控制方法、装置、电子设备、存储介质和产品 - Google Patents

机器人的控制方法、装置、电子设备、存储介质和产品

Info

Publication number
WO2025199736A1
WO2025199736A1 PCT/CN2024/083780 CN2024083780W WO2025199736A1 WO 2025199736 A1 WO2025199736 A1 WO 2025199736A1 CN 2024083780 W CN2024083780 W CN 2024083780W WO 2025199736 A1 WO2025199736 A1 WO 2025199736A1
Authority
WO
WIPO (PCT)
Prior art keywords
robots
conflict
robot
conversation
control method
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/CN2024/083780
Other languages
English (en)
French (fr)
Inventor
刘佳伟
金师兴
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Zitiao Network Technology Co Ltd
Original Assignee
Beijing Zitiao Network Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Zitiao Network Technology Co Ltd filed Critical Beijing Zitiao Network Technology Co Ltd
Priority to PCT/CN2024/083780 priority Critical patent/WO2025199736A1/zh
Priority to CN202480003555.3A priority patent/CN121152710A/zh
Publication of WO2025199736A1 publication Critical patent/WO2025199736A1/zh
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • BPERFORMING OPERATIONS; TRANSPORTING
    • B25HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25JMANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00Program-controlled manipulators
    • B25J9/16Program controls

Definitions

  • the present disclosure relates to the field of artificial intelligence technology, and in particular to a robot control method, device, electronic device, storage medium and product.
  • a robot control method comprising: in response to user input, obtaining a conversation of multiple robots with respect to the user's input, wherein the multiple robots are antagonistic with each other; performing conflict detection on the conversations of the multiple robots to determine the conversations in which conflicts exist and the types of conflicts; sending an adjustment instruction to at least one robot involved in the conflict so that each adjusted robot continues to have a conversation with the other robots based on the adjustment instruction, wherein the adjustment instruction is generated based on the user's input and the type of conflict; and displaying a response in response to the conversation of the multiple robots including a response with respect to the user's input and the absence of conflicts in the conversations of the multiple robots.
  • a robot control device comprising: an acquisition module configured to acquire, in response to a user's input, a dialogue of multiple robots with respect to the user's input, wherein the multiple robots are antagonistic with each other; a detection module configured to perform conflict detection on the dialogues of the multiple robots to determine the dialogues in which conflicts exist and the types of conflicts; a sending module configured to send an adjustment instruction to at least one robot involved in the conflict, so that each adjusted robot continues to dialogue with the other robots based on the adjustment instruction, the adjustment instruction being generated based on the user's input and the type of conflicts; and a display module configured to display a response in response to the dialogue of the multiple robots including a response to the user's input and the absence of conflicts in the dialogues of the multiple robots.
  • an electronic device comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the robot control method of any embodiment of the present disclosure based on instructions stored in the memory.
  • a computer-readable storage medium on which a computer program is stored.
  • the program is executed by a processor, the robot control method of any embodiment of the present disclosure is performed.
  • Users can input data through their device, for example, through an input interface provided by an application on the device.
  • User input can be a task assigned by the user to the robot, which can be represented by at least one of text, audio, images, and video.
  • User input involves multiple dimensions, including semantics and the processing flow used to process the input. For example, semantics can encompass multiple dimensions based on the literal meaning of the user's input, or it can be determined to encompass multiple dimensions through a correlation search and analysis of the topics covered by the user's input. For example, if the user input is "Please recommend me a mobile phone with good performance and low price," the literal semantics can reveal that it involves the dimensions of performance and price.
  • semantic analysis reveals that the topic is mobile phone recommendations.
  • the topic of mobile phone recommendations may involve multiple dimensions, such as performance, price, and appearance.
  • the robot in the embodiments of the present disclosure refers to an intelligent agent that can generate responses to input information, which can be implemented in the form of software, hardware, or a combination of software and hardware.
  • the robot can also be called a digital human or a virtual agent of a machine learning model.
  • the robot can be implemented based on a machine learning model, such as a large language model (Large Language Model).
  • a machine learning model can be implemented using a language model (LLM) or foundation model.
  • LLM language model
  • a machine learning model can be a generative model, which is used to output target content based on input information.
  • the input information of the generative model includes the processing basis of the generative model during the generation process, such as the information used to perform the generation process and the requirements for the target output content.
  • Generative models include, for example, models that generate data based on text or images, and the output of the generative model can include text, images, or a combination of the two.
  • the input or output of a generative model can also be data in other modalities, such as audio, video, or a combination of multiple types of data.
  • a generative model can be a single-modality model, such as a model that generates text from text (referred to as a "text-to-text model”) or a model that generates images from images (referred to as a "image-to-image model").
  • a generative model can be a cross-modal model, where the input and output belong to different modalities, such as a model that generates images from text (referred to as a "text-to-image model").
  • the input of a generative model can include multiple modalities, and the output can also include multiple modalities.
  • the robot processes it as needed and then feeds it into a machine learning model.
  • the robot then obtains the output of the machine learning model and processes it as needed to generate the robot's output.
  • the robot's output can be displayed directly on the user device's interface or transmitted to other modules in the user device (such as other robots or virtual objects) for further processing.
  • the adversarial nature of multiple robots manifests itself in their responses generated based on adversarial configurations. For example, these robots rely on different machine learning models trained with adversarial training data; or, when processing input, these inputs also include adversarial configurations. For example, in the scenario of recommending mobile phones to users, one robot might prioritize performance, while another might prioritize price.
  • each robot can first process the user's input and generate its own message (i.e., the message in the conversation). Alternatively, each robot can also process the user's input and the responses generated by other robots to generate its own response. Each robot's response can be visible to the user or transmitted only between robots. In other words, the "chat" between robots can be presented in a visual manner to show the user the game process between robots; or the conversation process between robots can be hidden, and only the response results after the robots reach a consensus are presented to the user.
  • step S104 conflict detection is performed on the dialogues of the multiple robots to determine conflicting dialogues and conflict types.
  • key information can be extracted.
  • This key information can reflect the key content of one or more messages from a single robot, or it can summarize the conversations between multiple robots, or it can be a key information sequence composed of multiple extracted key information.
  • the key information and key information sequence are then matched with pre-set conflict detection information to determine whether a conflict exists.
  • the conversations between robots can be directly matched with the conflict detection information to determine whether a conflict exists.
  • Other conflict detection methods can also be used by those skilled in the art as needed, and will not be described in detail here.
  • step S106 an adjustment instruction is sent to at least one robot involved in the conflict, so that each adjusted robot continues to communicate with other robots based on the adjustment instruction, where the adjustment instruction is generated based on the user input and the conflict type.
  • Adjustment instructions are used to notify the robot to adjust information such as message generation methods and content.
  • the robot can adjust its settings accordingly or temporarily add the adjustment instructions to its settings.
  • the robot can process the adjustment instructions to update the dialogue strategy. For example, the adjustment instructions or content determined based on the adjustment instructions can be input into a machine learning model to obtain the output of the machine learning model.
  • all or some of the robots involved in the conflicting conversation can be identified as robots to be adjusted. That is, adjustments can be made to all or some of the robots involved. For example, some robots can be instructed to make concessions, or all robots can be instructed to back down.
  • the dimensions involved in the conflict type can be determined, and the robots can be instructed to make opposing adjustments along those dimensions. For example, if robots are deadlocked over price, the robot insisting on a lower price can be instructed to raise its bottom line.
  • the adjustment instructions also include conflicting conversations. This more clearly identifies problematic conversations, helping the robot make more accurate adjustments. However, since each robot can receive messages from other robots and store its own messages, it is not necessary to send conflicting conversations to the robot again.
  • step S108 the dialogue of the plurality of robots includes a response to the user's input, and the plurality of robots There is no conflict in the conversation between the two robots, and the response is displayed.
  • a robot may engage in multiple rounds of dialogue, and the first round of dialogue may not necessarily fully respond to the user's input. For example, for a user-input task of recommending mobile phones, multiple robots may initially identify candidate models and then further narrow them down through discussion. Semantic analysis can be used to determine whether a robot's dialogue is a response to the user's input—that is, whether the robot has completed the communication and discussion related to the user's input.
  • the conflict detection method described in the above steps can be used to further determine whether there is a conflict. If there is no conflict, it means that the robots have reached an agreement, and the robot's response can be displayed to the user.
  • the above embodiment utilizes multiple robots with adversarial relationships to jointly develop responses to user inputs, enabling these responses to be considered across multiple dimensions and resulting in higher reliability.
  • conflicts may arise during the conversation.
  • adjustment instructions can be sent to at least one of the robots involved in the conflict, allowing the conversation between the robots to continue, thereby improving the efficiency of user responses.
  • user input involves multiple dimensions, and each of the multiple robots generates a conversation based on some or all of these dimensions.
  • Different robots may have overlapping dimensions, but their dimensions are not completely identical, creating a competitive environment between the robots.
  • Robot A's dimensions may include high performance and attractive appearance
  • Robot B's dimensions may include low price and attractive appearance. Both robots possess the dimension of "attractive appearance," but are competitive in terms of performance and price.
  • the conflict detection can be performed based on the dimensions involved in the robot's dialogue.
  • An embodiment of the conflict detection method disclosed herein is described below with reference to FIG2 .
  • Figure 2 shows a schematic flow chart of a conflict detection method according to some embodiments of the present disclosure. As shown in Figure 2 , the conflict detection method of this embodiment includes steps S202 to S204.
  • the dimensions corresponding to each robot in the multi-round dialogue can be analyzed as a whole based on the multi-round dialogue between robots. For example, a robot insists on a low price at the beginning of the dialogue, but makes concessions on the price during the multi-round interaction between the robots. In this case, the dialogue of the robot does not involve low prices. The dimension of price.
  • At least one of the topic of the conversation, the flow of the conversation, and the interaction pattern between the multiple robots is determined; and the dimensions involved in at least one of the topic of the conversation, the flow of the conversation, and the interaction pattern between the multiple robots are determined as the dimensions involved in the conversation between the multiple robots.
  • semantic analysis of the robot conversation can be performed to obtain any one of the topic, the flow of the conversation, and the interaction pattern between the robots, and then the dimension can be matched with a preset template for each dimension of the topic, the flow of the conversation, and the interaction pattern between the robots to determine the dimensions involved in the conversation.
  • step S204 in response to the fact that the dimensions involved in the conversation among the multiple robots include conflicting dimensions, it is determined that the conversation among the multiple robots has a conflict.
  • conflict type After determining the existence of a conflict, the conflict type can be further determined.
  • the conflict type of the conversation can be determined based on the type of dimension in which the conflict occurs.
  • conflict types can include topic conflicts, conversation flow conflicts, and conflicts in the interaction patterns between robots. These types can also be further subdivided into the above three types. Of course, conflict types can also be categorized in other ways, which will not be detailed here.
  • the dimensions involved in the robot's conversation can be extracted, and whether there is a conflict in the conversation can be determined based on the conflict between the dimensions. Therefore, conflict detection can be performed more accurately.
  • negative conversation examples of the conflict may also be predetermined to perform conflict detection based on the negative conversation examples.
  • a conflict detection method according to another embodiment of the present disclosure is described below with reference to FIG3 .
  • Figure 3 shows a schematic flow chart of a conflict detection method according to some other embodiments of the present disclosure. As shown in Figure 3 , the conflict detection method of this embodiment includes steps S302 to S304.
  • step S302 the conversations of the multiple robots are matched with negative conversation examples, where the negative conversation examples correspond to each of the multiple candidate conflict types.
  • Negative dialogue examples refer to conflicting dialogue examples, which can be pre-set before the robot's dialogue begins. Examples collected and set from other conversation scenarios, or obtained through other means. Negative conversation examples can be the conversation text itself or abstracted from negative conversation text. For example, negative conversation examples include "Robot T refuses to adjust its strategy even though it clearly conflicts with Robot U's goals" and "Robot T ignores the importance of shared resources, resulting in inefficiency.”
  • step S304 based on the matching results of the conversations of the multiple robots and the negative conversation examples, conflicting conversations and conflict types in the conversations between the multiple robots are determined.
  • the machine learning model can also be used for processing.
  • the conversation between the robots and the prompt are input into a conflict detection controller based on the machine learning model.
  • the prompt can include an instruction to perform conflict detection on the conversation between the robots, as well as negative conversation examples.
  • the machine learning model can process the conversation between the robots based on the prompt, and by referring to the negative conversation examples and leveraging the powerful semantic understanding and analysis capabilities of the machine learning model, efficiently and accurately detect conflicts and conflict types in the conversation.
  • conflict detection can be performed based on negative dialogue examples, which helps to accurately detect and analyze conflict information in detail, and improves the efficiency of subsequent conflict resolution.
  • all robots involved in the conflicting conversations can be identified as robots to be adjusted, or some robots involved in the conflicting conversations can be identified as robots to be adjusted.
  • some robots involved in the conflicting conversations can be further identified as robots to be adjusted based on user input, the conflict type, and the adjustment priorities of the multiple robots. Adjustment instructions can then be sent to the robots to be adjusted.
  • the robot to be adjusted can be determined based on the correlation between the robot's configuration information and the user's input. For example, if the user input is "Recommend me a mobile phone that is easy to use and preferably lasts longer," while this input involves both price and performance, the phrase "preferably lasts longer" indicates that the user prioritizes performance. Therefore, in the event of a conflict, the robot making recommendations based on price can be instructed to compromise and make concessions.
  • the conflict type is "information not shared".
  • the conflict is caused by robots that do not share information, so robot A, where the conflict is located, can be used as the robot to be adjusted.
  • Figure 4 shows a flow chart of a method for generating an adjustment instruction according to some embodiments of the present disclosure. As shown in Figure 4 , the method for generating an adjustment instruction in this embodiment includes steps S402 to S404.
  • step S402 positive historical conversation examples are identified from historical conversations between multiple robots that can be used to resolve conflicts of the conflict type. Specifically, solutions to similar conflicts encountered by multiple robots during the current task are used as a reference for the current conflict. Because positive historical conversation examples are generated by the current robot, adjustments to the robot causing the conflict can be made more effectively.
  • step S404 an adjustment instruction for the first robot is generated based on the task, the conflict type, and the positive dialogue examples and positive historical dialogue examples corresponding to the conflict type, where the adjustment instruction includes the positive historical dialogue example.
  • Adjustment instructions are used to instruct the robot how to adjust. They can be specific adjustment strategies, such as "making concessions on price” or “sharing known background information with other robots.” They can also be positive historical dialogue examples, positive dialogue examples, or both.
  • the above embodiment improves the robot's adjustment efficiency by using positive historical dialogue examples as a basis for adjustment when instructing the robot to adjust, thereby improving the efficiency of the robot's response to user input.
  • the conflict type includes a target inconsistency.
  • Generating an adjustment instruction includes: The system comprises the following steps: determining an adjustment information for the target of the first robot based on the input, the conflict type and an example corresponding to the conflict type, and the targets of the first robot and the second robot involved in the conflict; and generating an adjustment instruction for the first robot based on the adjustment information.
  • the number of the first and second robots can be one or more.
  • a bot's goal can be understood as something determined by the bot based on user input and its settings. For example, if the user input is "Recommend a phone for me,” and the bot prefers to recommend low-priced phones, the bot's initial goal would be "Recommend a phone for me, priced no more than xx.” This goal may be slightly adjusted as the conversation progresses.
  • the target adjustment information may include at least one of an adjustment direction and an adjustment amplitude, to instruct the first robot to compromise to partially achieve its original target; or, may include an adjusted target, to instruct the first robot to redefine the target.
  • the conflict type includes policy inconsistency.
  • Generating an adjustment instruction includes: generating a new policy based on the original policies of the multiple robots in response to the user input, the conflict type, and examples corresponding to the conflict type; and generating an adjustment instruction based on the new policy.
  • a policy refers to the operational strategy adopted by the robot in response to the user input. The actions performed based on the operational strategy are reflected as the flow of the conversation. For example, a policy can be a series of steps required to respond to the user, or information to be obtained or generated in different situations.
  • the conflict type includes a conflict in information understanding.
  • Generating an adjustment instruction includes: determining the background information involved in the conflicting conversation; based on the user's input, the conflict type, and an example corresponding to the conflict type, generating an instruction to provide background information, or an instruction to accept background information as an adjustment instruction.
  • Conflicts in information understanding include, for example, differences in information understanding or misunderstandings. Since different robots have different setting information, different robots may rely on different information sources or background knowledge when generating messages. Adjustment instructions can be provided to instruct the robots to merge their respective information in order to eliminate conflicts in information understanding. By providing an instruction for background information, the robot can actively send information that the other party does not have; by accepting an instruction for background information, the robot can regard the information provided by the other party as part of its own information.
  • the conflict type includes a resource allocation conflict.
  • Generating the adjustment instruction includes: generating details of setting information for the plurality of robots based on the tasks, the conflict type, and examples corresponding to the conflict type; generating a resource sharing strategy based on the details of the setting information for the plurality of robots; and generating the adjustment instruction based on the details of the setting information for the plurality of robots and the resource sharing strategy.
  • Resource conflicts and robot settings are often interrelated. For example, in resource allocation scenarios, different roles may encounter conflicts due to varying resource requirements. Differences in settings can also affect resource allocation and utilization. By referencing robot settings when generating resource sharing strategies and adjustment instructions, we can more comprehensively resolve robot dialogue conflicts.
  • the conversation process between multiple robots can be invisible or visible to the user. Adjustment instructions for the robots can be invisible or visible to the user. This is described below with reference to Figures 5A to 5C.
  • FIG. 5A shows a schematic diagram of a user interface according to some embodiments of the present disclosure.
  • the interface 51 includes a dialogue between a user 511 and an intelligent virtual object 512.
  • the user inputs "Please recommend a mobile phone to me.”
  • the background of the application where the interface 51 is located (which can be a terminal or a server) calls multiple adversarial robots 513, 514, and 515 to conduct a "dialogue" that is invisible to the user.
  • the conflict controller 516 located in the background performs conflict detection and sends adjustment instructions to the robots when a conflict is detected.
  • Figure 5B shows a schematic diagram of a user interface according to other embodiments of the present disclosure.
  • the interface 52 includes a conversation between a user 521 and multiple robots 523, 524, and 525, and the conversations between the robots are visible to the user.
  • the conflict controller 525 located in the background monitors the content of the conversations between the robots and performs conflict detection. According to the detection results, the conflict controller 525 found that the robot 523 was too insistent on the requirements for mobile phone performance, and did not give in at all to the recommendations of other robots, causing the conversation to reach a deadlock.
  • the conflict controller 525 can send an adjustment instruction to the robot 523 to relax its performance requirements.
  • the robots 523, 524, and 525 unanimously decided to recommend mobile phone Y of brand X, thereby responding to the user's input without conflict.
  • This method presents the discussion process between robots to users, which makes it easier for users to obtain more relevant information.
  • Figure 5C illustrates a user interface diagram according to yet another embodiment of the present disclosure.
  • interface 53 includes a conversation between user 531 and multiple robots 533 (Robot 1), 534 (Robot 2), and 535 (Robot 3).
  • the conversations between the robots are visible to the user.
  • conflict controller 535 as a participant in the group chat, sends a message within the group chat to issue adjustment instructions when a conflict is detected. Based on the detection results, conflict controller 535 finds that robot 533 is overly agreeent on its phone performance requirements and completely refuses to compromise on the recommendations of other robots, leading to a deadlock in the conversation.
  • Conflict controller 535 can send an instruction to robot 2 within the group chat to relax the performance standards, guiding the conversation smoothly until the three robots reach an agreement.
  • This approach presents the discussion and adjustment process between the robots to the user, making it easier for the user to obtain more relevant information and allowing the user to better understand the recommendation logic of each robot, providing further reference for the final decision.
  • FIG6 is a schematic diagram of a robot control device according to some embodiments of the present disclosure.
  • the robot control device 60 of this embodiment includes: an acquisition module 601 configured to, in response to user input, acquire a dialogue between multiple robots in response to the user input, wherein the multiple robots are antagonistic; a detection module 602 configured to perform conflict detection on the dialogues between the multiple robots to determine conflicting dialogues and conflict types; a sending module 603 configured to send an adjustment instruction to at least one robot involved in the conflict, so that each adjusted robot continues to communicate with the other robots based on the adjustment instruction, wherein the adjustment instruction is generated based on the user input and the conflict type; and a display module 604 configured to display the response in response to the dialogue between the multiple robots including a response to the user input and the absence of conflict in the dialogue between the multiple robots.
  • the user's input involves multiple dimensions, and each of the multiple robots generates a dialogue based on some or all of the multiple dimensions.
  • the detection module 602 is further configured to determine dimensions involved in the conversations among the multiple robots; and in response to the dimensions involved in the conversations among the multiple robots including conflicting dimensions, determine that the conversations among the multiple robots are in conflict.
  • the detection module 602 is further configured to determine at least one of the topic of the conversation, the flow of the conversation, and the interaction mode between the multiple robots based on the conversation between the multiple robots; and determine the dimensions involved in at least one of the topic of the conversation, the flow of the conversation, and the interaction mode between the multiple robots as the dimensions involved in the conversation between the multiple robots.
  • the detection module 602 is further configured to determine conflicting conversations and conflict types in the conversations between the multiple robots based on the matching results of the conversations of the multiple robots with negative conversation examples, wherein the negative conversation examples correspond to each of the multiple candidate conflict types.
  • the sending module 603 is further configured to: determine all robots involved in the conflicting conversation as robots to be adjusted, or, based on user input, conflict type, and adjustment priority of multiple robots, determine some robots from the robots involved in the conflicting conversation as robots to be adjusted; and send adjustment instructions to the robots to be adjusted.
  • control device 60 of the robot further includes a generation module 605 configured to generate an adjustment instruction.
  • the generation module 605 is further configured to generate an adjustment instruction based on the user's input, the conflict type, and a positive dialogue example corresponding to the conflict type.
  • the generation module 605 is further configured to determine positive historical dialogue examples for resolving conflicts of the conflict type in historical dialogues between multiple robots; based on the task, the conflict type, and the positive dialogue examples and positive historical dialogue examples corresponding to the conflict type, generate an adjustment indication for the first robot, the adjustment indication including the positive historical dialogue example.
  • the conflict type includes target inconsistency
  • the generation module 605 is further configured to determine adjustment information of the target of the first robot based on user input, the conflict type and an example corresponding to the conflict type, and the targets of the first robot and the second robot involved in the conflict; and generate adjustment instructions for the first robot based on the adjustment information.
  • the conflict type includes policy inconsistency
  • the generation module 605 is further configured to generate a new policy based on the original policies of multiple robots for user input, the conflict type, and examples corresponding to the conflict type; and generate an adjustment instruction based on the new policy.
  • the conflict type includes a conflict in information understanding
  • the generation module 605 is further configured to determine the background information involved in the conflicting conversation; based on the user input, the conflict type, and examples corresponding to the conflict type, an indication to provide background information or an indication to accept background information is generated as an adjustment indication.
  • the conflict type includes a conflict of resource allocation
  • the generation module 605 is further configured to generate details of setting information of multiple robots based on the task, conflict type and examples corresponding to the conflict type; generate a resource sharing strategy based on the details of the setting information of multiple robots; and generate adjustment instructions based on the details of the setting information of multiple robots and the resource sharing strategy.
  • the adjustment indication also includes the presence of conflicting conversations.
  • the above-mentioned units are merely logical modules divided according to the specific functions they implement, and are not intended to limit specific implementation methods. For example, they can be implemented in software, hardware, or a combination of software and hardware. In actual implementation, the above-mentioned units can be implemented as independent physical entities, or can also be implemented by a single entity (for example, a processor (CPU or DSP, etc.), an integrated circuit, etc.). In addition, the above-mentioned units are shown with dotted lines in the accompanying drawings to indicate that these units may not actually exist, and the operations/functions they implement can be implemented by the processing circuit itself.
  • the device may also include a memory that can store various information generated by the device and the various units contained in the device during operation, programs and data used for operation, data to be sent by the communication unit, etc.
  • the memory can be volatile memory and/or non-volatile memory.
  • the memory can include but is not limited to random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), read-only memory (ROM), and flash memory.
  • RAM random access memory
  • DRAM dynamic random access memory
  • SRAM static random access memory
  • ROM read-only memory
  • flash memory flash memory
  • the device may also include a communication unit that can be used to communicate with other devices.
  • the communication unit can be implemented in an appropriate manner known in the art, for example, including communication components such as an antenna array and/or a radio frequency link, various types of interfaces, communication units, etc. This will not be described in detail here.
  • the device may also include other components not shown, such as a radio frequency link, a baseband processing unit, a network interface, a processor, a controller, etc. This will not be described in detail here.
  • FIG. 7 shows a schematic structural diagram of an electronic device according to some embodiments of the present disclosure.
  • the electronic device 7 can be various types of devices, for example, including but not limited to mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc.
  • the electronic device 7 may include a display panel for displaying data and/or execution results utilized in the scheme of the present disclosure.
  • the display panel can be of various shapes, such as a rectangular panel, an elliptical panel, or a polygonal panel.
  • the display panel can be not only a flat panel, but also a curved panel or even a spherical panel.
  • the memory 71 is used to store one or more computer readable instructions.
  • the processor 72 is used to execute the computer readable instructions, the computer readable instructions are executed by the processor 72 according to any of the above embodiments.
  • the method described in the embodiment For the specific implementation of each step of the method and related explanations, please refer to the above embodiment, and the repeated parts will not be repeated here.
  • the processor 72 and the memory 71 may communicate with each other directly or indirectly.
  • the processor 72 and the memory 71 may communicate with each other via a network.
  • the network may include a wireless network, a wired network, and/or any combination of wireless networks and wired networks.
  • the processor 72 and the memory 71 may also communicate with each other via a system bus, which is not limited in this disclosure.
  • the processor 72 can be embodied as various appropriate processors, processing devices, etc., such as a central processing unit (CPU), a graphics processing unit (GPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • the central processing unit (CPU) can be an X86 or ARM architecture, etc.
  • the memory 71 can include any combination of various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory.
  • FIG8 shows a schematic structural diagram of a computer system according to some embodiments of the present disclosure.
  • a central processing unit (CPU) 801 performs various processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 to a random access memory (RAM) 803.
  • ROM read-only memory
  • RAM random access memory
  • the central processing unit is merely exemplary, and it may also be another type of processor, such as the various processors described above.
  • ROM 802, RAM 803, and storage portion 808 may be various forms of computer-readable storage media, as described below. It should be noted that although ROM 802, RAM 803, and storage device 808 are shown separately in FIG8 , one or more of them may be combined or located in the same or different memory or storage modules.
  • the CPU 801, ROM 802, and RAM 803 are connected to each other via a bus 804.
  • An input/output interface 805 is also connected to the bus 804.
  • the following components are connected to the input/output interface 805: an input portion 806 such as a touch screen, a touch pad, a keyboard,
  • the computer system 800 includes a computer program product 807, which includes a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), a speaker, a vibrator, and the like; a storage unit 808, which includes a hard disk, a magnetic tape, and the like; and a communication unit 809, which includes a network interface card such as a LAN card, a modem, and the like.
  • the communication unit 809 allows communication processing to be performed via a network such as the Internet.
  • FIG8 shows that the various devices or modules in the computer system 80 communicate via bus 804, they may also communicate via a network or other means, wherein the network may include a wireless network, a wired network, and/or any combination of wireless and wired networks.
  • Computer-readable storage media may include, but are not limited to, an electrical connection with one or more wires, 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), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.
  • a computer-readable storage medium may be any tangible medium that contains or stores a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.
  • a computer readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, in which a computer readable program code is carried.
  • the data signal transmitted may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof.
  • a computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.
  • Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.
  • the computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.
  • a computer program comprising: instructions, which, when executed by a processor, cause the processor to perform any of the methods of the above embodiments.
  • the instructions may be embodied as computer program codes.

Landscapes

  • Engineering & Computer Science (AREA)
  • Robotics (AREA)
  • Mechanical Engineering (AREA)
  • Manipulator (AREA)

Abstract

一种机器人的控制方法,包括:响应于用户的输入,获取多个机器人针对用户的输入的对话,其中,多个机器人之间具有对抗性;对多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型;向冲突涉及的至少一个机器人发送调整指示,以便每个被调整的机器人基于调整指示继续与其他机器人对话,调整指示是基于用户的输入和冲突类型生成的;响应于多个机器人的对话包括针对用户的输入的响应、并且多个机器人的对话不存在冲突,显示响应。还提供一种机器人的控制装置、电子设备、存储介质和产品。

Description

机器人的控制方法、装置、电子设备、存储介质和产品 技术领域
本公开涉及人工智能技术领域,特别涉及一种机器人的控制方法、装置、电子设备、存储介质和产品。
背景技术
由人工智能技术驱动的机器人可以基于接收的信息进行反馈。例如,在对话场景中,机器人可以根据其他发言者的对话来产生自己的对话。机器人往往依托于机器学习模型产生响应,这些模型可以使用特定领域的数据进行训练、以使得机器人能够针对特定领域的特点进行响应,或者机器学习模型在处理来自其他发言者的对话时,还参考机器人的设定信息,以产生满足该设定信息的对话。
发明内容
提供该发明内容部分以便以简要的形式介绍构思,这些构思将在后面的具体实施方式部分被详细描述。该发明内容部分并不旨在标识要求保护的技术方案的关键特征或必要特征,也不旨在用于限制所要求的保护的技术方案的范围。
根据本公开的一些实施例,提供了一种机器人的控制方法,包括:响应于用户的输入,获取多个机器人针对用户的输入的对话,其中,多个机器人之间具有对抗性;对多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型;向冲突涉及的至少一个机器人发送调整指示,以便每个被调整的机器人基于调整指示继续与其他机器人对话,调整指示是基于用户的输入和冲突类型生成的;响应于多个机器人的对话包括针对用户的输入的响应、并且多个机器人的对话不存在冲突,显示响应。
根据本公开的另一些实施例,提供了一种机器人的控制装置,包括:获取模块,被配置为响应于用户的输入,获取多个机器人针对用户的输入的对话,其中,多个机器人之间具有对抗性;检测模块,被配置为对多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型;发送模块,被配置为向冲突涉及的至少一个机器人发送调整指示,以便每个被调整的机器人基于调整指示继续与其他机器人对话,调整指示是基于用户的输入和冲突类型生成的;显示模块,被配置为响应于多个机器人的对话包括针对用户的输入的响应、并且多个机器人的对话不存在冲突,显示响应。
根据本公开的一些实施例,提供一种电子设备,包括:存储器;和耦接至存储器的处理器,处理器被配置为基于存储在存储器中的指令,执行本公开中的任一实施例的机器人的控制方法。
根据本公开的一些实施例,提供一种计算机可读存储介质,其上存储有计算机程序,该程序被处理器执行时执行本公开中的任一实施例的机器人的控制方法。
根据本公开的一些实施例,提供一种计算机程序产品,当计算机程序产品在计算机上运行时,使得计算机实现本公开中的任一实施例的机器人的控制方法。
根据本公开的一些实施例,提供一种计算机程序,包括:指令,指令当由处理器执行时使处理器执行本公开中的任一实施例的机器人的控制方法。
通过以下参照附图对本公开的示例性实施例的详细描述,本公开的其它特征、方面及其优点将会变得清楚。
附图说明
下面参照附图说明本公开的优选实施例。此处所说明的附图用来提供对本公开的进一步理解,各附图连同下面的具体描述一起包含在本说明书中并形成说明书的一部分,用于解释本公开。应当理解的是,下面描述中的附图仅仅涉及本公开的一些实施例,而非对本公开构成限制。在附图中:
图1示出了根据本公开一些实施例的机器人的控制方法的流程示意图。
图2示出了根据本公开一些实施例的冲突检测方法的流程示意图。
图3示出了根据本公开另一些实施例的冲突检测方法的流程示意图。
图4示出了根据本公开一些实施例的调整指示的生成方法的流程示意图。
图5A示出了根据本公开一些实施例的用户界面示意图。
图5B示出了根据本公开另一些实施例的用户界面示意图。
图5C示出了根据本公开又一些实施例的用户界面示意图。
图6示出了根据本公开一些实施例的机器人的控制装置的结构示意图。
图7示出了根据本公开一些实施例的的电子设备的结构示意图。
图8示出了根据本公开一些实施例的计算机系统的结构示意图。
应当明白,为了便于描述,附图中所示出的各个部分的尺寸并不一定是按照实际的比例关系绘制的。在各附图中使用了相同或相似的附图标记来表示相同或者相似的部件。因此,一旦某一项在一个附图中被定义,则在随后的附图中可能不再对其进行 进一步讨论。
具体实施方式
下面将结合本公开实施例中的附图,对本公开实施例中的技术方案进行清楚、完整地描述,但是显然,所描述的实施例仅仅是本公开一部分实施例,而不是全部的实施例。以下对实施例的描述实际上也仅仅是说明性的,决不作为对本公开及其应用或使用的任何限制。应当理解的是,本公开可以通过各种形式来实现,而且不应该被解释为限于这里阐述的实施例。
应当理解,本公开的方法实施方式中记载的各个步骤可以按照不同的顺序执行,和/或并行执行。此外,方法实施方式可以包括附加的步骤和/或省略执行示出的步骤。本公开的范围在此方面不受限制。除非另外具体说明,否则在这些实施例中阐述的部件和步骤的相对布置、数字表达式和数值应被解释为仅仅是示例性的,不限制本公开的范围。
本公开中使用的术语“包括”及其变型意指至少包括后面的元件/特征、但不排除其他元件/特征的开放性术语,即“包括但不限于”。此外,本公开使用的术语“包含”及其变型意指至少包含后面的元件/特征、但不排除其他元件/特征的开放性术语,即“包含但不限于”。因此,包括与包含是同义的。术语“基于”意指“至少部分地基于”。
整个说明书中所称“一个实施例”、“一些实施例”或“实施例”意味着与实施例结合描述的特定的特征、结构或特性被包括在本发明的至少一个实施例中。例如,术语“一个实施例”表示“至少一个实施例”;术语“另一实施例”表示“至少一个另外的实施例”;术语“一些实施例”表示“至少一些实施例”。而且,短语“在一个实施例中”、“在一些实施例中”或“在实施例中”在整个说明书中各个地方的出现不一定全都指的是同一个实施例,但是也可以指同一个实施例。
需要注意,本公开中提及的“第一”、“第二”等概念仅用于对不同的装置、模块或单元进行区分,并非用于限定这些装置、模块或单元所执行的功能的顺序或者相互依存关系。除非另有指定,否则“第一”、“第二”等概念并非意图暗示如此描述的对象必须按时间上、空间上、排名上的给定顺序或任何其他方式的给定顺序。
需要注意,本公开中提及的“一个”、“多个”的修饰是示意性而非限制性的,本领域技术人员应当理解,除非在上下文另有明确指出,否则应该理解为“一个或多 个”。
本公开实施方式中的多个装置之间所交互的消息或者信息的名称仅用于说明性的目的,而并不是用于对这些消息或信息的范围进行限制。
下面结合附图对本公开的实施例进行详细说明,但是本公开并不限于这些具体的实施例。下面这些具体实施例可以相互结合,对于相同或者相似的概念或过程可能在某些实施例不再赘述。此外,在一个或多个实施例中,特定的特征、结构或特性可以由本领域的普通技术人员从本公开将清楚的任何合适的方式组合。
由于机器人可以针对输入的信息生成对话,因此用户可以通过与机器人对话来获得信息,例如机器人可以作为搜索工具来向用户提供搜索结果、作为咨询工具来向用户提供某个任务的解决方案。然而,一个机器人所提供的内容可能具有局限性,使得向用户提供的响应不准确,降低了用户的信息获取效率。因此,本公开利用多个具有对抗性的机器人共同对用户的输入做出响应,以使得该响应能够兼顾多个机器人所参考的维度。下面参考图1描述本公开机器人的控制方法的实施例。
图1示出了根据本公开一些实施例的机器人的控制方法的流程示意图。如图1所示,该实施例的控制方法包括步骤S102至S108。
在步骤S102中,响应于用户的输入,获取多个机器人针对用户的输入的对话,其中,多个机器人之间具有对抗性。
用户可以通过用户设备输入,例如,在用户设备中的应用所提供的输入界面中输入。用户的输入可以是用户向机器人所部署的任务,其可以通过文本、音频、图像、视频中的至少一种表示。用户的输入涉及多个维度,多个维度可以是语义方面,也可以是在处理该输入的过程中所使用的处理流程方面。以语义为例,可以是用户的输入的字面意思包括多个维度,也可以是通过对用户的输入所涉及的主题进行关联搜索和分析后,确定其包括多个维度。例如,用户的输入为“请给我推荐一部性能好、价格便宜的手机”,则通过其字面的语义可以得到其涉及性能、价格两个维度;又例如,用户的输入为“请给我推荐一部手机”,通过分析其语义,得到其主题是推荐手机,通过搜索和分析与推荐手机相关的信息,可以得到推荐手机这一主题下可能涉及性能、价格、外观等多个维度。
本公开实施例中的机器人是指能够针对输入信息生成响应的智能体,其可以以软件、硬件或者软件和硬件结合的方式实现。机器人又可以被称为数字人、机器学习模型的虚拟代理。机器人可以依托于机器学习模型实现,例如基于大语言模型(Large  Language Model,简称:LLM)或者基础模型(Foundation Model)实现。机器学习模型可以是生成式模型,生成式模型用于基于输入的信息,输出目标内容。生成式模型输入的信息包括生成式模型在生成过程中的处理依据,例如参考哪些信息执行生成过程、输出的目标内容的要求等等。生成式模型例如包括基于文本进行生成的模型、或基于图像进行生成的模型,生成式模型的输出可以包括文本、图像或二者的结合。当然,生成式模型的输入或输出也可以是其他模态的数据,如音频、视频或者多种类型的数据的结合。生成式模型可以为单一模态的模型,例如基于文本生成文本的模型(简称“文生文模型”)、基于图像生成图像的模型(简称“图生图模型”);或者,生成式模型也可以是跨模态的模型,即输入和输出属于不同模态的模型,例如基于文本生成图像的模型(简称“文生图模型”);再或者,生成式模型的输入可以包括多个模态,输出也可以包括多个模态。
例如,机器人在接收到输入信息之后,根据需要对输入信息进行处理,再将输入信息输入到机器学习模型中;然后,机器人获取机器学习模型输出的结果,根据需要对模型输出的结果进行处理,以生成机器人的输出。机器人的输出可以直接显示在用户设备的界面,也可以传输给用户设备中的其他模块(例如其他机器人或者虚拟对象)继续处理。
多个机器人的对抗性体现在,多个机器人依据具有对抗性的设定信息,生成响应内容。例如,多个机器人依托不同的机器学习模型,这些机器学习模型是通过具有对抗性的训练数据训练的;又或者,多个机器人在处理输入信息时,输入信息还包括具有对抗性的设定信息。仍以向用户推荐手机的场景为例,该场景下,可以是一个机器人以性能优先进行推荐,另一个机器人以价格优先进行推荐。
在获得用户的输入后,每个机器人可以首先对用户的输入进行处理,并生成机器人的消息(即对话中的消息),或者,每个机器人还可以对用户的输入以及其他机器人生成的回复进行处理以生成自己的回复。每个机器人的回复可以是用户可见的,也可以是仅在机器人之间传输。即,可以以可视化的方式呈现机器人之间的“聊天”,以向用户展示机器人之间的博弈过程;也可以隐藏机器人之间的对话过程,仅向用户呈现最终各个机器人达成一致后的响应结果。
在步骤S104中,对多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型。
由于多个机器人之间具有对抗性,因此对个机器人之间的对话可能存在冲突。如 果该冲突始终得不到解决,那么机器人之间甚至可能陷入僵局,导致对话的实质内容没有进展。从而,用户无法及时有效地获得对用户的输入的响应。
在进行冲突检测时,可以在对机器人之间的对话进行语义分析后,提取关键信息,该关键信息可以是反映一个机器人的一条或多条消息中的关键内容,也可以是对多个机器人之间的对话的总结,还可以是将提取的多条关键信息构成关键信息序列。然后,将关键信息与关键信息序列与预置的冲突检测信息进行匹配,以识别是否存在冲突。或者,也可以直接将机器人之间的对话与冲突检测信息进行匹配,以识别是否存在冲突。根据需要,本领域技术人员也可以使用其他的冲突检测方法,这里不再赘述。
在一些实施例中,可以利用冲突控制器对机器人的对话进行监控。例如,向控制器输入用于冲突检测的提示信息(Prompt),以指示冲突控制器识别存在冲突的对话和冲突类型。提示信息中可以包括冲突检测信息。冲突检测信息例如包括负面对话示例以及其对应的冲突类型,负面对话示例为发生了冲突的对话的示例。
在步骤S106中,向冲突涉及的至少一个机器人发送调整指示,以便每个被调整的机器人基于调整指示继续与其他机器人对话,调整指示是基于用户的输入和冲突类型生成的。
调整指示用于通知机器人对消息生成方式、生成内容等生成信息进行调整,机器人可以据此调整其设定信息,也可以是在设定信息的基础上临时叠加该调整指示。机器人可以对调整指示进行处理以更新对话策略。例如,将调整指示或者基于调整指示确定的内容输入到机器学习模型中,以获得机器学习模型输出的内容。
在发生冲突时,可以将存在冲突的对话所涉及的所有机器人或者部分机器人确定为待调整的机器人。即,可以令冲突涉及的全部机器人进行调整,也可以令冲突涉及的部分机器人进行调整。例如,可以指示部分机器人进行让步,也可以指示全部机器人各退一步。在一些实施例中,可以确定冲突类型所涉及的维度,并指示机器人在该维度上进行相反的调整。例如,机器人在价格问题上产生冲突并僵持不下,此时可以指示其中坚持低价的机器人调高价格的底线。
在一些实施例中,调整指示还包括存在冲突的对话。从而可以更明确地指出哪些对话存在问题,有助于机器人更准确地进行调整。然而,由于每个机器人能够接收其他机器人发送的消息、以及能够存储自身的消息,因此也可以不将存在冲突的对话单独再发送给机器人一次。
在步骤S108中,响应于多个机器人的对话包括针对用户的输入的响应、并且多 个机器人的对话不存在冲突,显示响应。
机器人基于用户的输入可能产生多轮对话,并不一定首轮对话就能够完全响应用户的输入。例如,对于用户输入的推荐手机的任务,多个机器人可能先初步确定备选的型号,再进一步地通过讨论从中进行筛选。可以通过语义分析来确定机器人的对话是否是针对用户输入的响应,即,机器人是否完成了针对用户输入的沟通和讨论。
在机器人的对话包括针对用户的输入的响应的情况下,通过上述步骤描述的冲突检测的方式可以进一步判断是否存在冲突。如不存在,说明机器人之间已达成一致,可以将机器人的该响应显示给用户。
上述实施例利用了具有对抗性的多个机器人共同得出对用户输入的响应,使得该响应能够综合多个维度进行考量,具有更高的可靠度。然后,由于机器人之间存在对抗性,因此在对话的过程中可能产生冲突。在机器人的对话存在冲突、甚至陷入僵持的情况下,通过向冲突涉及的至少一个机器人发送调整指示,从而使得机器人之间的对话能够继续推进,提高了对用户响应的效率。
在一些实施例中,用户的输入涉及多个维度,多个机器人中的每一个基于多个维度中的部分或全部维度生成对话。不同的机器人可以有部分维度重叠,但是不同机器人的维度不完全相同,从而机器人之间具备对抗性。例如,机器人A的维度可以包括性能高、外观漂亮,机器人B的维度可以包括价格低、外观漂亮。二者都具备“外观漂亮”这一维度,但是在性能维度和价格维度上具备对抗性。
在进行冲突检测时,可以基于机器人的对话所涉及的维度进行冲突检测。下面参考图2描述本公开冲突检测方法的实施例。
图2示出了根据本公开一些实施例的冲突检测方法的流程示意图。如图2所示,该实施例的冲突检测方法包括步骤S202至S204。
在步骤S202中,确定多个机器人的对话所涉及的维度。例如,可以对机器人的对话进行语义分析,确定对话的关键信息,并将关键信息与用户的输入涉及的多个维度进行匹配,以确定对话所涉及的维度。
由于机器人之间存在对抗性,因此某个机器人发送的消息可能很容易和另外一个机器人发送的消息存在冲突的维度。但是,部分机器人在多轮交互的过程中会自行地调整。因此,在一些实施例中,可以基于机器人之间的多轮对话整体地分析每个机器人在这多轮对话中所对应的维度。例如,某个机器人在对话开始时坚持低价格,但是在机器人的多轮交互过程中对价格进行了让步,那么此时该机器人的对话则不涉及低 价格的维度。
在一些实施例中,基于多个机器人的对话,确定对话的主题、对话的流程、多个机器人之间的互动模式中的至少一项;将对话的主题、对话的流程、多个机器人之间的互动模式中的至少一项所涉及的维度,确定为多个机器人的对话所涉及的维度。例如,可以通过对机器人的对话进行语义分析,获得主题、对话的流程、机器人之间的互动模式中的任意一项,并将其与预设的主题、对话的流程、机器人之间的互动模式的各个维度的模板进行匹配,以确定对话所涉及的维度。
对话的主题是指通过对对话主题进行分析后所得的对话的核心内容。对话的流程是指对用户的输入进行分析和处理的流程,例如在为用户推荐手机时,可以先锁定较多数量的备选的手机型号、再从中选择,也可以先由每个机器人推举一款,再基于机器人之间的多轮对话寻找更合适的替代款。互动模式是指机器人是分工处理不同的任务,还是分组以由每组处理不同的任务,或者是集体处理某个任务,在对用户的输入进行处理的不同阶段可以使用不同的互动模式。在主题、对话的流程、机器人之间的互动模式中的任意一项产生冲突,都可能导致对话无法推进。
在步骤S204中,响应于多个机器人的对话所涉及的维度包括存在冲突的维度,确定多个机器人的对话存在冲突。
在确定存在冲突之后,可以进一步确定冲突的类型。例如,根据存在冲突的维度所属的类型,确定对话的冲突类型。冲突类型可以包括主题冲突、对话的流程冲突、机器人之间的互动模式的冲突,也可以进一步包括上述三种冲突之下的进一步细分。当然,还可以根据其他方式划分冲突类型,这里不再赘述。
通过上述实施例,可以通过对机器人的对话所涉及的维度进行提取,根据维度之间的冲突性确定对话中是否存在冲突。因此,能够更准确地进行冲突检测。
除了基于对话涉及的维度进行冲突检测之外,还可以预先确定冲突的负面对话示例,以基于负面对话示例进行冲突检测。下面参考图3描述本公开另一个实施例的冲突检测的方法。
图3示出了根据本公开另一些实施例的冲突检测方法的流程示意图。如图3所示,该实施例的冲突检测方法包括步骤S302至S304。
在步骤S302中,将多个机器人的对话与负面对话示例进行匹配,其中,负面对话示例与多个候选的冲突类型中的每一个对应。
负面对话示例是指存在冲突的对话示例,其可以是在机器人的对话开始之前预先 收集并设置的其他对话场景下的示例、或者是通过其他方式获得的示例。负面对话示例可以是对话文本本身,也可以是基于负面的对话文本抽象出的内容。例如,负面对话示例为“机器人T拒绝调整其策略,即使它与机器人U的目标明显冲突”“机器人T忽视共享资源的重要性,导致效率低下”。
在步骤S304中,基于多个机器人的对话与负面对话示例的匹配结果,确定多个机器人之间的对话中存在冲突的对话以及冲突类型。
在进行匹配时,可以将机器人的对话原文与负面对话示例进行匹配,这种方式的识别准确性比较高;或者,也可以对机器人的对话原文进行语义分析和抽象,提取出对话的主干内容,再与负面对话示例进行匹配,这种方式可以提高匹配的命中率。
在一些实施例中,也可以由机器学习模型进行处理。例如,将机器人之间的对话以及提示语(prompt)输入到基于机器学习模型的冲突检测控制器。提示语可以包括对机器人之间的对话进行冲突检测的指示,还包括负面对话示例。从而,机器学习模型可以基于提示语对机器人之间的对话进行处理,以通过参考负面对话示例,借助机器学习模型强大的语义理解和分析能力,高效、准确地检测对话中的冲突以及冲突类型。
通过上述实施例,可以基于负面对话示例进行冲突检测,有助于对冲突信息进行准确的检测和详细的分析,提高了后续冲突解决的效率。
如前文所述,在指示机器人进行调整时,可以将存在冲突的对话所涉及的所有机器人确定为待调整的机器人,也可以将将存在冲突的对话所涉及的部分机器人确定为待调整的机器人。在确定待调整的机器人时,可以进一步基于用户的输入、冲突类型、多个机器人的调整优先级,从存在冲突的对话所涉及的机器人中确定出部分机器人作为待调整的机器人。然后,可以向待调整的机器人发送调整指示。
基于用户的输入,可以确定出哪个机器人的设定信息(例如机器人所涉及的维度)对于用户的输入的相关性更大。即,可以根据机器人的设定信息与用户的输入的相关性,确定待调整的机器人。例如,用户的输入为“为我推荐一部偏移好用的手机,最好能用得久一点”,虽然该输入涉及价格和性能两个维度,但是通过“最好能用得久一点”,可以体现用户对性能的维度的偏重更高。从而在发生冲突的情况下,可以指示基于低价格进行推荐的机器人进行妥协和让步。
基于冲突类型,可以确定冲突的源头在哪一个机器人。例如,由于机器人A不共享信息而造成机器人A和B的冲突,冲突的类型为“信息不共享”,该类型的冲突由 不共享信息的机器人所引发,因此可以将冲突源头所在的机器人A作为待调整的机器人。
基于多个机器人的调整优先级,可以确定在冲突发生时,优先指示哪个机器人进行调整。该优先级的信息可以是预先设置的,也可以在机器人交互的过程中动态地指定。
通过使用以上各个方式确定待调整的机器人,可以更高效地提供解决冲突的可行方案,提高了冲突解决的效率,因此提高了响应用户的效率。
在一些实施例中,基于用户的输入、冲突类型以及与冲突类型对应的正面对话示例,生成调整指示。正面对话示例是指在对话过程中存在冲突、但是随着对话的推进解决了冲突的示例,其可以是在机器人的对话开始之前预先收集并设置的其他对话场景下的示例、或者是通过其他方式获得的示例。正面对话示例可以是对话文本本身,也可以是基于正面的对话文本抽象出的内容。通过正面对话示例,可以使得生成的调整指示具备更高的合理性和可行性。下面参考图4描述本公开调整指示的生成方法的实施例。
图4示出了根据本公开一些实施例的调整指示的生成方法的流程示意图。如图4所示,该实施例的调整指示的生成方法包括步骤S402至S404。
在步骤S402中,确定多个机器人之间的历史对话中,用于解决冲突类型的冲突的正面历史对话示例。即,通过多个机器人在处理当前任务时,对于类似冲突的解决方案供当前的冲突参考。由于正面历史对话示例是当前的机器人生成的,因此能够更有效地对产生冲突的机器人进行调整。
在步骤S404中,基于任务、冲突类型以及与冲突类型对应的正面对话示例和正面历史对话示例,生成对第一机器人的调整指示,调整指示包括正面历史对话示例。
调整指示用于指示机器人如何调整,其可以是具体的调整策略,例如“在价格维度上让步”“将已知的背景信息共享给其他机器人”,也可以是正面历史对话示例、正面对话示例,也可以兼而有之。
上述实施例通过在指示机器人调整时,将正面历史对话示例作为调整的依据,从而能够提高机器人的调整效率,因此对用户的输入的响应效率也得到了提高。
下面示例性地针对目标冲突、策略冲突、信息理解的冲突、资源分配的冲突这几种冲突类型,示例性地描述几种调整指示的生成方法。
在一些实施例中,冲突类型包括目标不一致。生成调整指示包括:基于用户的输 入、冲突类型以及与冲突类型对应的示例、冲突涉及的第一机器人和第二机器人的目标,确定第一机器人的目标的调整信息;基于调整信息,生成第一机器人的调整指示。第一和第二机器人的数量都可以为一个或多个。
机器人的目标可以理解为机器人基于用户的输入和机器人的设定信息确定。例如,用户的输入为“为我推荐一部手机”,而该机器人是倾向于推荐低价手机的机器人,则机器人的初始的目标为“为用户推荐一部手机,价格不超过xx”。随着对话的推进,该目标可能有轻微的调整。
目标的调整信息可以包括调整方向、调整幅度中的至少一项,以指示第一机器人进行妥协以部分实现自身原来的目标;或者,可以包括调整后的目标,以指示第一机器人重新确定目标。
该实施例提供了包括目标的调整信息的调整指示,来指示机器人直接解决发生的问题,以平衡不同的目标。通过首先分析对抗性数字人间的冲突点,然后提出具体的解决方案,能够减少机器人之间的冲突,提高了对用户的响应效率。
在一些实施例中,冲突类型包括策略不一致。生成调整指示包括:基于多个机器人针对用户的输入的原始策略、冲突类型以及与冲突类型对应的示例,生成新的策略;基于新的策略,生成调整指示。策略是指机器人为响应用户的输入而采取的操作策略,基于操作策略执行的动作被体现为对话的流程。例如,策略可以是在响应用户时所需采取的一系列步骤,或者在不同情况下需要获取或生成的信息等等。
在机器人的策略产生冲突时,通过生成新的策略并指示机器人执行,能够对机器人之间的策略进行统一,从而能够减少机器人之间的冲突,提高了对用户的响应效率。
在一些实施例中,冲突类型包括信息理解的冲突。生成调整指示包括:确定存在冲突的对话所涉及的背景信息;基于用户的输入、冲突类型以及与冲突类型对应的示例,生成提供背景信息的指示、或者接受背景信息的指示作为调整指示。信息理解的冲突例如包括信息理解的差异或者误解。由于不同机器人的设定信息不相同,因此不同的机器人在生成消息时所依据的信息源、或者背景知识可能不同。调整指示能够提供用于指示机器人融合各自的信息,以便消除信息理解的冲突。通过提供背景信息的指示,能够使得机器人主动发送对方不具备的信息;通过接受背景信息的指示,能够使得机器人将对方提供的信息作为自身具备的信息的一部分。
通过促进知识融合,能够指示不同的机器人整合彼此的信息和观点,从而促进每个机器人做出更全面的决策,提高了对用户的响应的可靠度和可用性。
在一些实施例中,冲突类型包括资源分配的冲突。生成调整指示包括:基于任务、冲突类型以及与冲突类型对应的示例,生成多个机器人的设定信息的详情;基于多个机器人的设定信息的详情,生成资源共享策略;基于多个机器人的设定信息的详情、资源共享策略,生成调整指示。
资源冲突和机器人的设定信息往往是相互关联的。例如,在资源分配的场景中,不同角色可能会因为资源需求的不同而产生冲突。设定信息的不同也可能影响资源的分配和使用。在生成资源共享策略和调整指示时,通过参考机器人的设定信息,能够更全面地解决机器人的对话冲突。
因此,通过基于冲突类型生成机器人的设定信息的详情,能够在机器人的原始设定信息的基础上,进一步进行细化,以帮助明确和优化不同机器人的角色和职责。在此基础上,通过生成资源共享策略,能够兼顾资源分配的公平性和有效性,从整体上提高机器人的响应生成效率。
在本公开的一些实施例中,多个机器人之间的对话过程可以是用户不可见的、也可以是用户可见的。对机器人的调整指示可以是用户不可见的,也可以是用户可见的。下面示例性地参照图5A至5C进行描述。
图5A示出了根据本公开一些实施例的用户界面示意图。如图5A所示,界面51中包括用户511与智能虚拟对象512之间的对话。用户输入“请给我推荐一部手机”。在接收到用户的输入后,该界面51所在应用的后台(可以是终端或者服务器)调用多个具有对抗性的机器人513、514、515进行用户不可见的“对话”,在“对话”过程中,由位于后台的冲突控制器516进行冲突检测,并在检测到冲突时向机器人发送调整指示。在机器人53、54、55达成一致后,通过前台的智能虚拟对象52向用户反馈结果“X品牌的手机Y不错”。这种方式可以直接向用户反馈结果,提高了用户接收有效信息的效率。
图5B示出了根据本公开另一些实施例的用户界面示意图。如图5B所示,界面52中包括用户521与多个机器人523、524、525的对话,机器人之间的对话对用户可见。同时,位于后台的冲突控制器525监控机器人之间的对话内容并进行冲突检测。根据检测结果,冲突控制器525发现机器人523过于坚持对手机性能的要求,而对其他机器人的推荐完全不让步,使得对话陷入僵局。冲突控制器525可以向机器人523发送调整指示,使其放宽对性能的要求。再经过一些轮次的对话和讨论后,机器人523、524、525一致决定推荐X品牌的手机Y,从而对用户的输入进行了无冲突的响应。这 种方式将机器人之间的讨论过程呈现给用户,可以便于用户获取更多的相关信息。
图5C示出了根据本公开又一些实施例的用户界面示意图。如图5C所示,界面53中包括用户531与多个机器人533(机器人1)、534(机器人2)、535(机器人3)的对话,机器人之间的对话对用户可见。同时,冲突控制器535也作为参与群聊的一员,在检测到冲突发送时,通过在群聊中发送消息以发送调整指示。根据检测结果,冲突控制器535发现机器人533过于坚持对手机性能的要求,而对其他机器人的推荐完全不让步,使得对话陷入僵局。冲突控制器535可以在群聊中向机器人2发送“对性能的标准放宽一些”的指示,以引导对话顺利地进行,直到三个机器人达成一致。这种方式将机器人之间的讨论过程以及调整过程呈现给用户,可以便于用户获取更多的相关信息,也能够使得用户更清楚各个机器人的推荐逻辑,从而对最终的决策提供进一步的参考。
以上对本公开的一些实施例的机器人的控制方法进行了描述。下面结合其他附图描述本公开的相关装置的实施例。
图6示出了根据本公开一些实施例的机器人的控制装置的结构示意图。如图6所示,该实施例的机器人的控制装置60包括:获取模块601,被配置为响应于用户的输入,获取多个机器人针对用户的输入的对话,其中,多个机器人之间具有对抗性;检测模块602,被配置为对多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型;发送模块603,被配置为向冲突涉及的至少一个机器人发送调整指示,以便每个被调整的机器人基于调整指示继续与其他机器人对话,调整指示是基于用户的输入和冲突类型生成的;显示模块604,被配置为响应于多个机器人的对话包括针对用户的输入的响应、并且多个机器人的对话不存在冲突,显示响应。
在一些实施例中,用户的输入涉及多个维度,多个机器人中的每一个基于多个维度中的部分或全部维度生成对话。
在一些实施例中,检测模块602进一步被配置为确定多个机器人的对话所涉及的维度;响应于多个机器人的对话所涉及的维度包括存在冲突的维度,确定多个机器人的对话存在冲突。
在一些实施例中,检测模块602进一步被配置为基于多个机器人的对话,确定对话的主题、对话的流程、多个机器人之间的互动模式中的至少一项;将对话的主题、对话的流程、多个机器人之间的互动模式中的至少一项所涉及的维度,确定为多个机器人的对话所涉及的维度。
在一些实施例中,检测模块602进一步被配置为基于多个机器人的对话与负面对话示例的匹配结果,确定多个机器人之间的对话中存在冲突的对话以及冲突类型,其中,负面对话示例与多个候选的冲突类型中的每一个对应。
在一些实施例中,发送模块603进一步被配置为:将存在冲突的对话所涉及的所有机器人确定为待调整的机器人,或者,基于用户的输入、冲突类型、多个机器人的调整优先级,从存在冲突的对话所涉及的机器人中确定出部分机器人作为待调整的机器人;向待调整的机器人发送调整指示。
在一些实施例中,机器人的控制装置60还包括生成模块605,被配置为生成调整指示。
在一些实施例中,生成模块605进一步被配置为基于用户的输入、冲突类型以及与冲突类型对应的正面对话示例,生成调整指示。
在一些实施例中,生成模块605进一步被配置为确定多个机器人之间的历史对话中,用于解决冲突类型的冲突的正面历史对话示例;基于任务、冲突类型以及与冲突类型对应的正面对话示例和正面历史对话示例,生成对第一机器人的调整指示,调整指示包括正面历史对话示例。
在一些实施例中,冲突类型包括目标不一致,生成模块605进一步被配置为基于用户的输入、冲突类型以及与冲突类型对应的示例、冲突涉及的第一机器人和第二机器人的目标,确定第一机器人的目标的调整信息;基于调整信息,生成第一机器人的调整指示。
在一些实施例中,冲突类型包括策略不一致,生成模块605进一步被配置为基于多个机器人针对用户的输入的原始策略、冲突类型以及与冲突类型对应的示例,生成新的策略;基于新的策略,生成调整指示。
在一些实施例中,冲突类型包括信息理解的冲突,生成模块605进一步被配置为确定存在冲突的对话所涉及的背景信息;基于用户的输入、冲突类型以及与冲突类型对应的示例,生成提供背景信息的指示、或者接受背景信息的指示作为调整指示。
在一些实施例中,冲突类型包括资源分配的冲突,生成模块605进一步被配置为基于任务、冲突类型以及与冲突类型对应的示例,生成多个机器人的设定信息的详情;基于多个机器人的设定信息的详情,生成资源共享策略;基于多个机器人的设定信息的详情、资源共享策略,生成调整指示。
在一些实施例中,调整指示还包括存在冲突的对话。
应注意,上述各个单元仅是根据其所实现的具体功能划分的逻辑模块,而不是用于限制具体的实现方式,例如可以以软件、硬件或者软硬件结合的方式来实现。在实际实现时,上述各个单元可被实现为独立的物理实体,或者也可由单个实体(例如,处理器(CPU或DSP等)、集成电路等)来实现。此外,上述各个单元在附图中用虚线示出指示这些单元可以并不实际存在,而它们所实现的操作/功能可由处理电路本身来实现。
此外,尽管未示出,该设备也可以包括存储器,其可以存储由设备、设备所包含的各个单元在操作中产生的各种信息、用于操作的程序和数据、将由通信单元发送的数据等。存储器可以是易失性存储器和/或非易失性存储器。例如,存储器可以包括但不限于随机存储存储器(RAM)、动态随机存储存储器(DRAM)、静态随机存取存储器(SRAM)、只读存储器(ROM)、闪存存储器。当然,存储器可也位于该设备之外。可选地,尽管未示出,但是该设备也可以包括通信单元,其可用于与其它装置进行通信。在一个示例中,通信单元可以被按照本领域已知的适当方式来实现,例如包括天线阵列和/或射频链路等通信部件,各种类型的接口、通信单元等等。这里将不再详细描述。此外,设备还可以包括未示出的其它部件,诸如射频链路、基带处理单元、网络接口、处理器、控制器等。这里将不再详细描述。
本公开的一些实施例还提供一种电子设备。图7示出了根据本公开一些实施例的的电子设备的结构示意图。例如,在一些实施例中,电子设备7可以为各种类型的设备,例如可以包括但不限于诸如移动电话、笔记本电脑、数字广播接收器、PDA(个人数字助理)、PAD(平板电脑)、PMP(便携式多媒体播放器)、车载终端(例如车载导航终端)等等的移动终端以及诸如数字TV、台式计算机等等的固定终端。例如,电子设备7可以包括显示面板,以用于显示根据本公开的方案中所利用的数据和/或执行结果。例如,显示面板可以为各种形状,例如矩形面板、椭圆形面板或多边形面板等。另外,显示面板不仅可以为平面面板,也可以为曲面面板,甚至球面面板。
如图7所示,该实施例的电子设备7包括:存储器71以及耦接至该存储器71的处理器72。应当注意,图7所示的电子设备7的组件只是示例性的,而非限制性的,根据实际应用需要,该电子设备7还可以具有其他组件。处理器72可以控制电子设备7中的其它组件以执行期望的功能。
在一些实施例中,存储器71用于存储一个或多个计算机可读指令。处理器72用于运行计算机可读指令时,计算机可读指令被处理器72运行时实现根据上述任一实 施例所述的方法。关于该方法的各个步骤的具体实现以及相关解释内容可以参见上述的实施例,重复之处在此不作赘述。
例如,处理器72和存储器71之间可以直接或间接地互相通信。例如,处理器72和存储器71可以通过网络进行通信。网络可以包括无线网络、有线网络、和/或无线网络和有线网络的任意组合。处理器72和存储器71之间也可以通过系统总线实现相互通信,本公开对此不作限制。
例如,处理器72可以体现为各种适当的处理器、处理装置等,诸如中央处理器(CPU)、图形处理器(Graphics Processing Unit,GPU)、网络处理器(NP)等;还可以是数字信号处理器(DSP)、专用集成电路(ASIC)、现场可编程门阵列(FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。中央处理元(CPU)可以为X86或ARM架构等。例如,存储器71可以包括各种形式的计算机可读存储介质的任意组合,例如易失性存储器和/或非易失性存储器。存储器71例如可以包括系统存储器,系统存储器例如存储有操作系统、应用程序、引导装载程序(Boot Loader)、数据库以及其他程序等。在存储介质中还可以存储各种应用程序和各种数据等。
另外,根据本公开的一些实施例,根据本公开的各种操作/处理在通过软件和/或固件实现的情况下,可从存储介质或网络向具有专用硬件结构的计算机系统,例如图8所示的计算机系统80安装构成该软件的程序,该计算机系统在安装有各种程序时,能够执行各种功能,包括诸如前文所述的功能等等。图8示出了根据本公开一些实施例的计算机系统的结构示意图。
在图8中,中央处理单元(CPU)801根据只读存储器(ROM)802中存储的程序或从存储部分808加载到随机存取存储器(RAM)803的程序执行各种处理。在RAM 803中,也根据需要存储当CPU 801执行各种处理等时所需的数据。中央处理单元仅仅是示例性的,其也可以是其它类型的处理器,诸如前文所述的各种处理器。ROM 802、RAM 803和存储部分808可以是各种形式的计算机可读存储介质,如下文所述。需要注意的是,虽然图8中分别示出了ROM 802、RAM 803和存储装置808,但是它们中的一个或多个可以合并或者位于相同或不同的存储器或存储模块中。
CPU 801、ROM 802和RAM 803经由总线804彼此连接。输入/输出接口805也连接到总线804。
下述部件连接到输入/输出接口805:输入部分806,诸如触摸屏、触摸板、键盘、 鼠标、图像传感器、麦克风、加速度计、陀螺仪等;输出部分807,包括显示器,比如阴极射线管(CRT)、液晶显示器(LCD),扬声器,振动器等;存储部分808,包括硬盘,磁带等;和通信部分809,包括网络接口卡比如LAN卡、调制解调器等。通信部分809允许经由网络比如因特网执行通信处理。容易理解的是,虽然图8中示出计算机系统80中的各个装置或模块是通过总线804来通信的,但它们也可以通过网络或其它方式进行通信,其中,网络可以包括无线网络、有线网络、和/或无线网络和有线网络的任意组合。
根据需要,驱动器810也连接到输入/输出接口805。可拆卸介质811比如磁盘、光盘、磁光盘、半导体存储器等等根据需要被安装在驱动器810上,使得从中读出的计算机程序根据需要被安装到存储部分808中。
在通过软件实现上述系列处理的情况下,可以从网络比如因特网或存储介质比如可拆卸介质811安装构成软件的程序。
根据本公开的实施例,上文参考流程图描述的过程可以被实现为计算机软件程序。例如,本公开的实施例包括一种计算机程序产品,其包括承载在计算机可读介质上的计算机程序,该计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中,该计算机程序可以通过通信装置809从网络上被下载和安装,或者从存储装置808被安装,或者从ROM 802被安装。在该计算机程序被CPU 801执行时,执行本公开实施例的方法中限定的上述功能。
需要说明的是,在本公开的上下文中,计算机可读介质可以是有形的介质,其可以包含或存储以供指令执行系统、装置或设备使用或与指令执行系统、装置或设备结合地使用的程序。计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是,但不限于:电、磁、光、电磁、红外线、或半导体的系统、装置或器件,或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于:具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中,计算机可读存储介质可以是任何包含或存储程序的有形介质,该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。而在本公开中,计算机可读信号介质可以包括在基带中或者作为载波一部分传播的数据信号,其中承载了计算机可读的程序代码。这种传 播的数据信号可以采用多种形式,包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读信号介质还可以是计算机可读存储介质以外的任何计算机可读介质,该计算机可读信号介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输,包括但不限于:电线、光缆、RF(射频)等等,或者上述的任意合适的组合。
上述计算机可读介质可以是上述电子设备中所包含的;也可以是单独存在,而未装配入该电子设备中。
在一些实施例中,还提供了一种计算机程序,包括:指令,指令当由处理器执行时使处理器执行上述任一个实施例的方法。例如,指令可以体现为计算机程序代码。
在本公开的实施例中,可以以一种或多种程序设计语言或其组合来编写用于执行本公开的操作的计算机程序代码,上述程序设计语言包括但不限于面向对象的程序设计语言,诸如Java、Smalltalk、C++,还包括常规的过程式程序设计语言,诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中,远程计算机可以通过任意种类的网络(包括局域网(LAN)或广域网(WAN))连接到用户计算机,或者,可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。
附图中的流程图和框图,图示了按照本公开各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上,流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分,该模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意,在有些作为替换的实现中,方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如,两个接连地表示的方框实际上可以基本并行地执行,它们有时也可以按相反的顺序执行,这依所涉及的功能而定。也要注意的是,框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合,可以用执行规定的功能或操作的专用的基于硬件的系统来实现,或者可以用专用硬件与计算机指令的组合来实现。
描述于本公开实施例中所涉及到的模块、部件或单元可以通过软件的方式实现,也可以通过硬件的方式来实现。其中,模块、部件或单元的名称在某种情况下并不构 成对该模块、部件或单元本身的限定。
本文中以上描述的功能可以至少部分地由一个或多个硬件逻辑部件来执行。例如,非限制性地,可以使用的示例性的硬件逻辑部件包括:现场可编程门阵列(FPGA)、专用集成电路(ASIC)、专用标准产品(ASSP)、片上系统(SOC)、复杂可编程逻辑设备(CPLD)等等。
以上描述仅为本公开的一些实施例以及对所运用技术原理的说明。本领域技术人员应当理解,本公开中所涉及的公开范围,并不限于上述技术特征的特定组合而成的技术方案,同时也应涵盖在不脱离上述公开构思的情况下,由上述技术特征或其等同特征进行任意组合而形成的其它技术方案。例如上述特征与本公开中公开的(但不限于)具有类似功能的技术特征进行互相替换而形成的技术方案。
在本文提供的描述中,阐述了许多特定细节。然而,理解的是,可以在没有这些特定细节的情况下实施本发明的实施例。在其他情况下,为了不模糊该描述的理解,没有对众所周知的方法、结构和技术进行详细展示。
此外,虽然采用特定次序描绘了各操作,但是这不应当理解为要求这些操作以所示出的特定次序或以顺序次序执行来执行。在一定环境下,多任务和并行处理可能是有利的。同样地,虽然在上面论述中包含了若干具体实现细节,但是这些不应当被解释为对本公开的范围的限制。在单独的实施例的上下文中描述的某些特征还可以组合地实现在单个实施例中。相反地,在单个实施例的上下文中描述的各种特征也可以单独地或以任何合适的子组合的方式实现在多个实施例中。
虽然已经通过示例对本公开的一些特定实施例进行了详细说明,但是本领域的技术人员应该理解,以上示例仅是为了进行说明,而不是为了限制本公开的范围。本领域的技术人员应该理解,可在不脱离本公开的范围和精神的情况下,对以上实施例进行修改。本公开的范围由所附权利要求来限定。

Claims (17)

  1. 一种机器人的控制方法,包括:
    响应于用户的输入,获取多个机器人针对所述用户的输入的对话,其中,所述多个机器人之间具有对抗性;
    对所述多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型;
    向所述冲突涉及的至少一个机器人发送调整指示,以便每个被调整的机器人基于所述调整指示继续与其他机器人对话,所述调整指示是基于所述用户的输入和所述冲突类型生成的;
    响应于所述多个机器人的对话包括针对所述用户的输入的响应、并且所述多个机器人的对话不存在冲突,显示所述响应。
  2. 根据权利要求1所述的控制方法,其中,所述用户的输入涉及多个维度,所述多个机器人中的每一个基于所述多个维度中的部分或全部维度生成对话。
  3. 根据权利要求1或2所述的控制方法,其中,所述对所述多个机器人的对话进行冲突检测包括:
    确定所述多个机器人的对话所涉及的维度;
    响应于所述多个机器人的对话所涉及的维度包括存在冲突的维度,确定所述多个机器人的对话存在冲突。
  4. 根据权利要求3所述的控制方法,其中,所述确定所述多个机器人的对话所涉及的维度包括:
    基于所述多个机器人的对话,确定所述对话的主题、所述对话的流程、所述多个机器人之间的互动模式中的至少一项;
    将所述对话的主题、所述对话的流程、所述多个机器人之间的互动模式中的至少一项所涉及的维度,确定为所述多个机器人的对话所涉及的维度。
  5. 根据权利要求1至4中任一项所述的控制方法,其中,所述对所述多个机器人的对话进行冲突检测,以确定存在冲突的对话和冲突类型包括:
    基于所述多个机器人的对话与负面对话示例的匹配结果,确定所述多个机器人之间的对话中所述存在冲突的对话以及所述冲突类型,其中,所述负面对话示例与多个候选的冲突类型中的每一个对应。
  6. 根据权利要求1至5中任一项所述的控制方法,其中,所述向所述冲突涉及的至少一个机器人发送调整指示包括:
    将存在冲突的对话所涉及的所有机器人确定为待调整的机器人,或者,基于所述用户的输入、所述冲突类型、所述多个机器人的调整优先级,从所述存在冲突的对话所涉及的机器人中确定出部分机器人作为待调整的机器人;
    向所述待调整的机器人发送调整指示。
  7. 根据权利要求1至6中任一项所述的控制方法,还包括:
    基于所述用户的输入、所述冲突类型以及与所述冲突类型对应的正面对话示例,生成所述调整指示。
  8. 根据权利要求7所述的控制方法,其中,所述生成所述调整指示包括:
    确定所述多个机器人之间的历史对话中,用于解决所述冲突类型的冲突的正面历史对话示例;
    基于所述任务、所述冲突类型以及与所述冲突类型对应的正面对话示例和所述正面历史对话示例,生成对所述第一机器人的调整指示,所述调整指示包括所述正面历史对话示例。
  9. 根据权利要求1至8中任一项所述的控制方法,其中,所述冲突类型包括目标不一致,所述控制方法还包括:
    基于所述用户的输入、所述冲突类型以及与所述冲突类型对应的示例、所述冲突涉及的第一机器人和第二机器人的目标,确定所述第一机器人的目标的调整信息;
    基于所述调整信息,生成所述第一机器人的调整指示。
  10. 根据权利要求1至9中任一项所述的控制方法,其中,所述冲突类型包括策略不一致,所述控制方法还包括:
    基于所述多个机器人针对所述用户的输入的原始策略、所述冲突类型以及与所述冲突类型对应的示例,生成新的策略;
    基于所述新的策略,生成所述调整指示。
  11. 根据权利要求1至10中任一项所述的控制方法,其中,所述冲突类型包括信息理解的冲突,所述控制方法还包括:
    确定所述存在冲突的对话所涉及的背景信息;
    基于所述用户的输入、所述冲突类型以及与所述冲突类型对应的示例,生成提供所述背景信息的指示、或者接受所述背景信息的指示作为所述调整指示。
  12. 根据权利要求1至11中任一项所述的控制方法,其中,所述冲突类型包括资源分配的冲突,所述控制方法还包括:
    基于所述任务、所述冲突类型以及与所述冲突类型对应的示例,生成所述多个机器人的设定信息的详情;
    基于所述多个机器人的设定信息的详情,生成资源共享策略;
    基于所述多个机器人的设定信息的详情、所述资源共享策略,生成所述调整指示。
  13. 根据权利要求1至12中任一项所述的控制方法,其中,所述调整指示还包括所述存在冲突的对话。
  14. 一种机器人的控制装置,包括:
    存储器;以及
    耦接至所述存储器的处理器,所述处理器被配置为基于存储在所述存储器中的指令,执行如权利要求1至13中任一项所述的机器人的控制方法。
  15. 一种计算机可读存储介质,其上存储有计算机程序,该程序被处理器执行时实现权利要求1至13中任一项所述的机器人的控制方法。
  16. 一种计算机程序产品,当所述计算机程序产品在计算机上运行时,使得所述计算机实现权利要求1至13中任一项所述的机器人的控制方法。
  17. 一种计算机程序,包括:
    指令,所述指令当由处理器执行时使所述处理器执行根据权利要求1至13中任一项所述的机器人的控制方法。
PCT/CN2024/083780 2024-03-26 2024-03-26 机器人的控制方法、装置、电子设备、存储介质和产品 Pending WO2025199736A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2024/083780 WO2025199736A1 (zh) 2024-03-26 2024-03-26 机器人的控制方法、装置、电子设备、存储介质和产品
CN202480003555.3A CN121152710A (zh) 2024-03-26 2024-03-26 机器人的控制方法、装置、电子设备、存储介质和产品

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2024/083780 WO2025199736A1 (zh) 2024-03-26 2024-03-26 机器人的控制方法、装置、电子设备、存储介质和产品

Publications (1)

Publication Number Publication Date
WO2025199736A1 true WO2025199736A1 (zh) 2025-10-02

Family

ID=97217440

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/083780 Pending WO2025199736A1 (zh) 2024-03-26 2024-03-26 机器人的控制方法、装置、电子设备、存储介质和产品

Country Status (2)

Country Link
CN (1) CN121152710A (zh)
WO (1) WO2025199736A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190126488A1 (en) * 2017-11-02 2019-05-02 Hitachi, Ltd. Robot dialogue system and control method of robot dialogue system
US20190248019A1 (en) * 2018-02-15 2019-08-15 DMAI, Inc. System and method for dynamic robot profile configurations based on user interactions
CN111191000A (zh) * 2019-12-20 2020-05-22 北京淇瑀信息科技有限公司 一种智能语音机器人的对话管理方法、装置和系统
CN112035650A (zh) * 2020-09-07 2020-12-04 贝壳技术有限公司 信息处理方法、装置、计算机可读存储介质及电子设备
CN112634684A (zh) * 2020-12-11 2021-04-09 深圳市木愚科技有限公司 智能教学方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190126488A1 (en) * 2017-11-02 2019-05-02 Hitachi, Ltd. Robot dialogue system and control method of robot dialogue system
US20190248019A1 (en) * 2018-02-15 2019-08-15 DMAI, Inc. System and method for dynamic robot profile configurations based on user interactions
CN111191000A (zh) * 2019-12-20 2020-05-22 北京淇瑀信息科技有限公司 一种智能语音机器人的对话管理方法、装置和系统
CN112035650A (zh) * 2020-09-07 2020-12-04 贝壳技术有限公司 信息处理方法、装置、计算机可读存储介质及电子设备
CN112634684A (zh) * 2020-12-11 2021-04-09 深圳市木愚科技有限公司 智能教学方法和装置

Also Published As

Publication number Publication date
CN121152710A (zh) 2025-12-16

Similar Documents

Publication Publication Date Title
CN113377520B (zh) 资源调度方法、装置、设备以及存储介质
KR102862560B1 (ko) 딥러닝 모델 기반의 데이터 생성 방법, 트레이닝 방법 및 장치
CN112071302B (zh) 计算代理的合成语音选择
US20230385350A1 (en) Feedback method and apparatus based on online document comment, and non-transitory computer-readable storage medium
JP7121052B2 (ja) イメージデータに少なくとも部分的に基づく、アクションを実行するためのエージェントの決定
US12400129B2 (en) Method of knowledge sharing among dialogue systems, dialogue method and device
WO2025161295A1 (zh) 搜索方法、装置、计算机可读存储介质和计算机程序产品
EP4252149A1 (en) Method and system for over-prediction in neural networks
US20250199830A1 (en) Information display method, medium and electronic device
EP4134885A2 (en) Method and apparatus of recommending data, electronic device, and medium
WO2025222374A1 (zh) 交互方法、装置、电子设备、存储介质和产品
CN120353909A (zh) 融合不同大模型的智能体问答方法、装置、介质及产品
US20250021574A1 (en) Interaction method and apparatus, and electronic device and storage medium
CN112463973B (zh) 医学知识图谱的构建方法、装置、介质及电子设备
US20250337706A1 (en) Multimedia content recommendation method and device, electronic device and storage medium
US12229182B2 (en) Image-based search method, server, terminal, and medium
WO2025199736A1 (zh) 机器人的控制方法、装置、电子设备、存储介质和产品
US20260099520A1 (en) Conversation methods, electronic devices, storage media, and products
WO2025145892A1 (zh) 交互信息的生成方法、系统和电子设备
WO2025156919A1 (zh) 任务执行方法、系统、电子设备、存储介质和程序产品
WO2025161306A1 (zh) 互动信息展示方法、装置、设备及存储介质
CN111597311A (zh) 用于输出信息的方法和装置
CN117992675A (zh) 内容推荐方法、装置、电子设备及存储介质
CN118069917A (zh) 查询信息重写的方法和电子设备
CN116522012A (zh) 用户兴趣挖掘方法、系统、电子设备以及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24931625

Country of ref document: EP

Kind code of ref document: A1