Service robot and collective intelligence-based multi-robot collaboration method using rag-based service policies updated with information collected by the service robot

Abstract

The multi-robot collaboration method of the present invention may include the steps of: at least one service robot circulating a preset assigned area within a store; the at least one service robot collecting store information by exploring the assigned area; the at least one service robot updating store operation data corresponding to the store information by analyzing the store information; the at least one service robot converting the store operation data into a RAG-based service policy; the at least one service robot uploading the RAG-based service policy to a main server; and the at least one service robot accessing the RAG-based service policy to determine a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy.

Description

Technology Field

[0001] The present disclosure relates to a multi-robot collaboration method, and more specifically, to a collective intelligence-based multi-robot collaboration method utilizing a service robot and a RAG-based service policy updated with information collected by the service robot. Background Technology

[0002] In the field of service robot technology, robots are being utilized in retail stores, exhibition spaces, and public information areas to provide customers with location, product, and event information. For instance, service robots can respond to customer inquiries while moving through an entire store or specific areas. Additionally, service robots can perform guidance services based on information, operating rules, and FAQ data pre-entered into a central server or internal storage. Furthermore, in a multi-robot environment, multiple service robots can be deployed in different locations or zones to perform their assigned roles and tasks.

[0003] Meanwhile, various limiting factors exist in the actual operating environment of robot systems, and problems may arise due to these limitations. For example, in a large service space, it is difficult to propagate various new information generated locally or information that changes in real time to the entire space outside of that specific area; consequently, this can lead to issues where service operation policies are not reflected in a timely manner. Additionally, individual service robots may face limitations in immediately recognizing information that has not been pre-entered by a manager or operator. In such cases, service disparities may occur between robots as multiple service robots provide different guidance, response standards, and priority judgments regarding the same customer inquiry or operational situation. Furthermore, service disparities may occur within the same store depending on the location, as the latest operational information collected from a specific location or zone is not shared in a timely manner with service robots deployed in other locations or zones.

[0004] Therefore, to enhance operational accuracy and response speed in a service robot collaboration environment, technology is required that enables each service robot to actively collect information within its assigned area and facilitates real-time information sharing so that multiple service robots can utilize the collected data together. In particular, rather than maintaining information held by individual service robots in a separate state, a collective intelligence-based collaboration structure is needed where multiple service robots jointly reflect on-site information to provide consistent guidance and response. Furthermore, the central server must not be limited to merely storing or distributing information; it needs to function as a collaboration hub that integrates and coordinates data collected from each service robot to reflect it in overall service operations. In other words, technology is required that allows multiple service robots to share common decision criteria and up-to-date information while responding to real-time changing on-site conditions. Prior art literature

[0005] Korean Registered Patent No. 10-2771536 B1 “Method for Operating an Unmanned Clothing Store Using Object Recognition Artificial Intelligence and Autonomous Unmanned Robot” The problem to be solved

[0006] One objective of the present disclosure is to provide a service robot and a collective intelligence-based multi-robot collaboration method that can more accurately and sensitively reflect the real-time operational status of a store site by having each service robot repeatedly circulate its assigned area while collecting both interactive and non-interactive store information, and updating store operation data by reflecting the reliability of the complex store information.

[0007] Another objective of the present disclosure is to provide a service robot and a collective intelligence-based multi-robot collaboration method that can convert field information acquired by individual service robots into a common policy asset that all service robots can jointly utilize by converting store operation data into a RAG-based service policy reconstructed into searchable knowledge information, application condition information, and service strategy information, and accumulating the RAG-based service policy on a main server.

[0008] Another objective of the present disclosure is to provide a service robot and a collective intelligence-based multi-robot collaboration method in which each service robot accesses a RAG-based service policy of a main server to dynamically determine a circulating strategy, response strategy, interaction strategy, work strategy, and collaboration strategy, thereby responding in real time to changes in the store environment and customer demands, and enabling all multi-robots to perform cooperative and efficient services based on the same up-to-date policy.

[0009] However, the problems that this disclosure aims to solve are not limited to those mentioned above, and other unmentioned problems will be clearly understood by a person skilled in the art from the description below. means of solving the problem

[0010] A multi-robot collaboration method according to the present disclosure for achieving the above-described technical problem may include the steps of: at least one service robot circulating a preset assigned area within a store; the at least one service robot collecting store information by exploring the assigned area; the at least one service robot updating store operation data corresponding to the store information by analyzing the store information; the at least one service robot converting the store operation data into a RAG-based service policy; the at least one service robot uploading the RAG-based service policy to a main server; and the at least one service robot accessing the RAG-based service policy to determine a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy.

[0011] In one embodiment, in the step of circulating the preset assigned area, the at least one service robot may generate a movement path based on path information of the assigned area that includes at least a portion of the entire store area, and may repeatedly circulate the assigned area based on a circulating schedule, movement priority, stopping point, re-search cycle, and avoidance path corresponding to the assigned area.

[0012] In one embodiment, in the step of collecting the store information, the at least one service robot may actively query a store manager within the assigned area and, based on the response to the query, collect interactive store information including at least one of event information, discount information, out-of-stock information, restocking schedule information, congestion information, detour information, and customer guidance priority information.

[0013] In one embodiment, in the step of collecting the store information, the at least one service robot can collect non-interactive store information by detecting the shelf status, product display density, whether an event banner is exposed, customer dwell density, whether a queue is formed, whether passage is possible, and the staff placement status of the area in charge.

[0014] In one embodiment, in the step of collecting the store information, the at least one service robot can generate composite store information for the same assigned area by combining interactive store information and non-interactive store information, and calculate the reliability score C of the composite store information using the following [Formula 1].

[0015] [Formula 1]

[0016] C = w1M + w2V + w3X + w4R

[0017] Here, M is a value indicating whether the store manager has directly verified the data, V is a value indicating whether the interactive collection result and the non-interactive collection result match, X is a value indicating whether there is cross-verification among multiple service robots, and R may be a value indicating a recency score based on the collection time.

[0018] In one embodiment, in the step of updating the store operation data, the at least one service robot can analyze the store information by performing at least one of text conversion of voice recognition data, classification of object recognition data, mapping with location information, determination of event occurrence, linking of time information, identification of assigned area, and determination of priority.

[0019] In one embodiment, in the step of updating the store operation data, the at least one service robot extracts operation information from the store information including at least one of product location information, event schedule information, discount application information, out-of-stock information, restocking schedule information, congested area information, detour requirement information, customer precautions information, manager delivery information, and guidance priority information, and when a preset update condition is satisfied, the store operation data can be updated by comparing the operation information with previously stored store operation data.

[0020] In one embodiment, the RAG-based service policy may include knowledge information searchable when determining a customer query or a service to be provided, application condition information serving as a criterion for determining whether to provide a service or the method of providing a service, and the service strategy defining the execution direction for each service strategy. In the step of converting the store operation data into the RAG-based service policy, the at least one service robot may generate the knowledge information, the application condition information, and the service strategy by classifying the store operation data into entity information, assigned area information, time information, query type information, and strategy type information, and by recombining the classified store operation data based on information directly used for customer response, information determining whether to provide a service, and information used for performing a service.

[0021] In one embodiment, in the step of uploading the RAG-based service policy to the main server, the at least one service robot may upload the RAG-based service policy to the main server based on at least one of a customer urgent request, identification of out-of-stock status, exceeding of a standard congestion level, a preset cycle, and a store manager command, and the main server may store assigned area identification information, collection time information, service robot identification information, policy version information, update reason information, and reliability information together with the RAG-based service policy.

[0022] In one embodiment, the at least one service robot can determine the circulating strategy by accessing the RAG-based service policy to search for a policy that reflects the status of the assigned area, congestion, event status, and the operation status of other service robots, and by determining at least one of the re-inspection frequency, bypass area, movement path, and circulating priority area based on the priority score and reliability score of the searched policy.

[0023] In one embodiment, the at least one service robot can determine the response strategy by searching for the RAG-based service policy including a query type tag corresponding to a customer query, and determining the response content based on at least one of a response template identifier, priority score, product location information, discount application information, out-of-stock information, congestion area information, and detour need information included in the searched policy.

[0024] In one embodiment, the at least one service robot can determine the response strategy by searching for the RAG-based service policy corresponding to whether a customer is approaching, whether a manager is requesting, time zone, congestion level, and customer type, and determining at least one of whether to provide preemptive guidance, priority of response to a customer or manager, level of explanation, and tone of voice for response based on the searched policy.

[0025] In one embodiment, the at least one service robot can determine the work strategy by accessing the RAG-based service policy to search for a policy that reflects at least one of information discrepancy, lack of reliability, lack of stock, change in event status, and need for manager verification, and by determining the need for at least one of re-search, re-verification, store manager query, event banner verification, and repeated collection based on the searched policy.

[0026] In one embodiment, the at least one service robot can determine the cooperation strategy by accessing the RAG-based service policy to search for a policy that reflects the status of the assigned area, congestion status, response status, support request information, and customer handover requirement information of another service robot, and by determining at least one of role sharing, information re-verification request, adjacent area support request, customer handover, and temporary joint patrol based on the searched policy.

[0027] A service robot according to the present disclosure for achieving the technical objectives described above may include a memory for storing at least one instruction and at least one processor for executing said at least one instruction. The at least one processor may be configured to circulate a pre-set assigned area within a store, to collect store information by searching said assigned area, to update store operation data corresponding to said store information by analyzing said store information, to convert said store operation data into a RAG-based service policy, to upload said RAG-based service policy to a main server, and to determine a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy by accessing said RAG-based service policy.

[0028] In addition to this, a computer program stored on a computer-readable recording medium for implementing the present disclosure may be further provided.

[0029] In addition to this, a computer-readable recording medium for recording a computer program for implementing the present disclosure may be further provided. Effects of the invention

[0030] According to the aforementioned means for solving the problem of the present disclosure, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure allows each service robot to repeatedly circulate its assigned area to collect both conversational and non-conversational store information, and to update store operation data by reflecting the reliability of the complex store information. Since the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure may include a structure that considers whether there is a direct response from an administrator, whether the information matches, whether there is cross-verification, and the recency of the information, it can reduce the problem of incorrect information or temporary detection results being directly reflected in the operation data. Accordingly, the service robot and the collective intelligence-based multi-robot collaboration method of the present disclosure, which utilizes a RAG-based service policy updated with information collected by the service robot, can improve service quality tailored to the actual store situation.

[0031] Furthermore, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure converts store operation data into a RAG-based service policy reconstructed into searchable knowledge information, application condition information, and service strategy information, and accumulates the said RAG-based service policy on a main server; thus, field information acquired by individual service robots can be converted into a common policy asset that can be jointly utilized by all service robots. The service robot and collective intelligence-based multi-robot collaboration method of the present disclosure can store assigned area identification information, collection time information, service robot identification information, policy version information, update reason information, and reliability information together on the main server, thereby enabling traceable management of the policy's source, recency, and scope of application. Accordingly, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure can perform consistent guidance and decision-making anytime and anywhere within the store.

[0032] Furthermore, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure allows each service robot to access the RAG-based service policy of the main server and dynamically determine traversal strategies, response strategies, interaction strategies, work strategies, and collaboration strategies, thereby enabling the performance of services with higher situational adaptability than fixed rule-based guidance. The service robot and collective intelligence-based multi-robot collaboration method of the present disclosure can adjust movement paths, response content, response priority, the need for reconfirmation, and the support status of adjacent robots by reflecting the latest policies such as event information, congestion information, out-of-stock information, detour information, and administrator request information, thus enabling real-time response to changes in the store environment and customer demands. Accordingly, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure enables all multi-robots to perform cooperative and efficient services based on the same latest service policy.

[0033] The effects of the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below. Brief explanation of the drawing

[0034] FIG. 1 is a block diagram illustrating the multi-robot collaboration method of the present invention. FIG. 2 is a diagram illustrating, for example, the areas in charge of multiple service robots within the entire store area. FIG. 3 is a block diagram illustrating the configuration of a service robot of the present disclosure. Figure 4 is a flowchart illustrating the detailed steps of the multi-robot collaboration method of the present invention. FIG. 5 is a diagram illustrating the operation of a service robot of the present invention collecting store information. FIG. 6 is a diagram illustrating the operation of a service robot of the present invention uploading an updated RAG-based service policy. FIG. 7 is a diagram illustrating the operation of a service robot of the present invention determining a service strategy based on RAG. FIG. 8 is a diagram showing an example of a service robot of the present invention determining a circulating strategy based on RAG. FIG. 9 is a diagram showing an example of a service robot of the present invention determining a response strategy based on RAG. FIG. 10 is a diagram showing an example of a service robot of the present invention determining a response strategy based on RAG. FIG. 11 is a diagram showing an example of a service robot of the present invention determining a work strategy based on RAG. FIG. 12 is a diagram showing an example of a service robot of the present invention determining a cooperation strategy based on RAG. Specific details for implementing the invention

[0035] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, various modifications may be made to the embodiments, and thus the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, and substitutions to the embodiments are included within the scope of the rights.

[0036] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification includes modifications, equivalents, or substitutions that fall within the technical concept.

[0037] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0038] Furthermore, terms defined in commonly used dictionaries are not interpreted ideally or excessively unless explicitly and specifically defined otherwise. In certain cases, terms have been selected at the applicant's discretion, and in such cases, their meanings will be described in detail in the relevant explanatory sections. Accordingly, terms used in this disclosure should be defined not merely by their names, but based on their meanings and the content throughout this disclosure.

[0039] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0040] The terms used in the embodiments are for illustrative purposes only and should not be interpreted as intended to be limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0041] Throughout this specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, the singular form used in this specification includes the plural form unless specifically stated otherwise in the text. Additionally, the expression "at least one of a, b, and c" described throughout this specification may encompass 'a alone', 'b alone', 'c alone', 'a and b', 'a and c', 'b and c', or 'a, b, and c all'.

[0042] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments pertain. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0043] Additionally, terms such as “part,” “module,” etc., as described in this specification refer to a unit that processes at least one function or operation, which may be implemented in hardware or software, or a combination of hardware and software. Furthermore, embodiments of this disclosure may be represented by functional block configurations and various processing steps. These functional blocks may be implemented by various numbers of hardware and / or software configurations that execute specific functions. For example, embodiments of this disclosure may employ integrated circuit configurations such as memory, processing, logic, look-up tables, etc., which can execute various functions under the control of one or more microprocessors or other control devices.

[0044] Each block of the process flow diagrams attached to this specification and combinations of the flow diagrams may be executed by computer program instructions. Since these computer program instructions may be loaded into the processor of a general-purpose computer, a computer for special purposes, or other programmable data processing equipment, the instructions executed through the processor of the computer or other programmable data processing equipment create means for performing the functions described in the flow diagram block(s).

[0045] These computer program instructions may be stored in computer-available or computer-readable memory that can be directed toward a computer or other programmable data processing equipment to implement a function in a specific way, and the instructions stored in said computer-available or computer-readable memory may also produce a manufactured item containing instruction means that performs the function described in the flowchart block(s).

[0046] Since computer program instructions can be loaded onto a computer or other programmable data processing equipment, instructions that perform a series of operation steps on the computer or other programmable data processing equipment to create a process executed by the computer can also provide steps for executing the functions described in the flowchart block(s).

[0047] Additionally, each block may represent a module, segment, or part of code containing one or more executable instructions for executing a specified logical function(s). Furthermore, in some alternative execution examples, the functions mentioned in the blocks may occur out of order. For instance, two blocks described in succession may actually be executed substantially simultaneously, or the blocks may be executed in reverse order according to their corresponding functions.

[0048] In addition, when describing with reference to the attached drawings, identical components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. In describing the embodiments, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the embodiments, such detailed description is omitted.

[0049] FIG. 1 is a block diagram showing a multi-robot collaboration method (100) of the present invention.

[0050] Referring to FIG. 1, the multi-robot collaboration method of the present invention can update a service policy using store information (120) collected by each of the multiple service robots (110) in the store. The multi-robot collaboration method can form a RAG (130) with the store information (120) collected by the multiple service robots (110) and the updated service policy information. The RAG (130) can provide functions for saving, modifying, deleting, and updating by linking with a main server (140).

[0051] Specifically, the multi-robot collaboration method of the present invention allows a plurality of service robots (110) to divide and manage different areas within a store, and store information obtained from each area can be collected into common policy information and utilized for service provision. For example, the service robots (110) of the present invention may be deployed in multiple numbers, such as a first service robot (110a), a second service robot (110b), and a third service robot (110c), and each service robot (110) can perform customer guidance and information collection in its assigned area.

[0052] The store information (120) of the present invention may include on-site operational status such as event information, schedule information, guidance information, out-of-stock information, and movement path information. That is, the present invention may be implemented not in a way where a single robot processes all information alone, but in a way where information collected by a plurality of service robots (110) is shared and used together through collective intelligence.

[0053] The service robot (110) illustrated in FIG. 1 is not merely a guidance device that simply provides store information, but can operate as an intelligent guidance entity that actually moves within the store and checks on-site changes. For example, each service robot (110) of the present invention can obtain new store information by conversing with a store manager or detecting surrounding conditions in its assigned area. Each service robot (110) of the present invention can transmit the obtained information in a form that can be reflected in the RAG (130) rather than storing it only internally. In the structure of the present invention, discount information confirmed by one service robot (110) in an event area can be utilized for customer responses by another service robot (110). In the structure of the present invention, congestion information confirmed by one service robot (110) near the checkout counter can also be utilized for movement decisions by another service robot (110).

[0054] The RAG (130) of the present invention may be a policy storage area where RAG-based service policies are stored. The RAG (130) may be a core component of the present invention for collecting and organizing store information collected by a plurality of service robots (110) and utilizing it as a common service policy. For example, the RAG (130) of the present invention may not be a storage area that simply stores store information, but may manage information in a policy structure that can be directly utilized to answer customer inquiries or determine the actions of service robots (110).

[0055] The main server (140) of the present invention is connected to the RAG (130) to manage overall policy information and maintain common standards that each service robot (110) can use. For example, the main server (140) of the present invention can manage information such as which information was collected in which area, what information it is at what point in time, and which service robot (110) transmitted the information. Multiple service robots (110) can share the same latest policy through the main server (140) while operating individually according to the situation of their respective assigned areas.

[0056] As shown in FIG. 1, the multi-robot collaboration method of the present invention may have a cyclic structure in which store information (120) is collected by a service robot (110), the collected information is organized into a common policy through a RAG (130) and a main server (140), and the organized policy is reflected in the guidance and movement of each service robot (110). The information / policy cyclic structure of the present invention can rapidly spread newly generated information and policies from the field to all robots. In the present invention, the experience and collected information of one service robot (110) can serve as a basis for judgment regarding the provision of services by other service robots (110). The cyclic structure of the present invention can provide more accurate guidance to customers and can enhance real-time response performance, such as avoiding crowded areas, providing out-of-stock notifications, and providing event notifications. Therefore, the multi-robot collaboration method of the present invention can provide a collective intelligence-based robot service that allows each service robot (110) to flexibly respond to changes in the field while operating the entire store as a single common knowledge system.

[0057] FIG. 2 is a drawing for exemplarily illustrating the areas in charge of a plurality of service robots in the entire store area (200).

[0058] Referring to FIG. 2, the entire store area (200) may be formed as a single continuous service space. Additionally, the entire store area (200) may be divided into multiple designated areas. For example, as illustrated in FIG. 2, the entire store area (200) may be divided into multiple designated areas such as area 1, area 2, area 3, and area 4, which are delineated by dotted lines. Each designated area does not simply mean a space with an evenly divided area, but may be a management unit that reflects operational characteristics such as product display locations, customer dwelling locations, table arrangements, checkout or guidance points, and locations accessible to store managers.

[0059] Multiple service robots (210a, 210b, 210c, 210d) may be distributed across the entire store area (200), and each service robot (210a, 210b, 210c, 210d) may be configured to prioritize different assigned areas. For example, the first service robot (210a) may operate primarily in the area corresponding to area 1. For example, the second service robot (210b) may operate primarily in the area corresponding to area 2. For example, the third service robot (210c) may operate primarily in the area corresponding to area 3. For example, the fourth service robot (210d) may operate primarily in the area corresponding to area 4. Each service robot may prioritize circulating its assigned area, but may perform supplementary search by moving to the vicinity of the boundary of an adjacent assigned area as needed.

[0060] The entire store area (200) may include a table area, a display shelf area, a checkout or work-related area, and a customer passageway. The objects of observation and collection may be set differently for each assigned area. For example, an area adjacent to a counter or workbench, such as area 1 or area 3, may be an area favorable for collecting interactive information with a store manager or employee. An area adjacent to a display shelf or customer table, such as area 2 or area 4, may be an area favorable for checking product condition, whether promotional banners are displayed, customer dwell density, and congestion status.

[0061] Multiple service robots (210a, 210b, 210c, 210d) are distributed by assigned zones, thereby shortening the information collection cycle for the entire store area (200). Each service robot can prioritize checking at least one of the following in its assigned zone: changes in product location, out-of-stock status, restocking schedule, whether there is a discount event, customer density, and whether passage is possible. Each service robot can later reflect the information collected in its assigned zone as common policy information, thereby enabling it to be used for guidance and judgment by other service robots. A structure in which multiple service robots repeatedly circulate in the same manner allows for faster identification of on-site changes in each zone while maintaining the entire store as a single common service space.

[0062] The layout structure illustrated in FIG. 2 is merely an exemplary environment for understanding the description of the present invention below and does not limit the store structure and assigned areas of the present invention. For example, the number, shape, size, boundary location of each assigned area within the store, and the initial placement location of each service robot may be varied depending on the structure of the store and the purpose of operation.

[0063] FIG. 3 is a block diagram for explaining a service robot according to an embodiment of the present disclosure.

[0064] Referring to FIG. 3, the service robot (300) may include a manipulator (310), at least one sensor (320), a main body (330), a driving unit (340), a display (350), a lift (360), at least one processor (370), a memory (380), and a communication unit (390).

[0065] According to one embodiment, the manipulator (310) may include a gripper (311). According to one embodiment, the gripper (311) may grasp or support a guidance terminal, a guidance card, a small item, a sample product, or a medium for verification by a store manager. For example, the gripper (311) may stably maintain a screen or input terminal that the store manager checks during interaction with the store manager, and may perform actions such as pointing to a location of guidance, delivering a small guide material, or pointing to a specific product or display stand during customer interaction. According to one embodiment, the gripper (311) may include at least one jaw, and the at least one jaw may fix the guidance terminal or item to reduce shaking during movement. According to one embodiment, the gripper (311) may stably maintain an object using an adsorption method, a friction method, an elastic support method, or an electrical fixing method. According to one embodiment, the manipulator (310) is not limited to a simple grasping function and may also be used for actions such as adjusting the sensor viewpoint, adjusting the screen direction during conversation with the store manager, or directing the display screen toward the customer.

[0066] According to one embodiment, at least one sensor (320) may include at least one of a first sensor (320-1) or a second sensor (320-2). For example, the first sensor (320-1) may include a depth camera including a time of flight (ToF) camera sensor, etc. According to one embodiment, the first sensing data obtained through the first sensor (320-1) may include depth information. The depth information may include depth values ​​corresponding to each of a plurality of pixels. For example, the ToF camera sensor may irradiate a signal, such as near-infrared rays, ultrasound, lasers, etc., and receive a reflected signal when the irradiated signal is reflected by a subject, such as a customer, store manager, display stand, display goods, aisle obstacles, event structures, etc. At least one processor (370) may measure the elapsed time from when the ToF camera sensor irradiates a signal until it receives a reflected signal to measure the distance or depth between the ToF camera sensor and the subject.

[0067] According to one embodiment, at least one processor (370) can identify a customer, store manager, display stand, displayed goods, aisle obstacle, or event banner adjacent to the service robot (300) based on first sensing data acquired through the first sensor (320-1), and can identify the size, distance, location, or occupancy status of the subject. For example, at least one processor (370) can identify a subject located within the field of view (FOV) of the first sensor (320-1) based on the first sensing data, for example, a customer, store manager, or display stand located in front of the service robot (300). However, it is not limited thereto, and the first sensor (320-1) may be implemented as various types of sensors capable of identifying the subject. For example, the first sensor (320-1) may include a stereo vision camera comprising at least two cameras. According to one embodiment, at least one processor (370) can identify the depth and size of a subject based on first sensing data that reflects binocular parallax characteristics in which the subject is captured differently through a stereo vision camera. For example, the stereo vision camera can acquire a left eye image and a right eye image by capturing the subject from different perspectives using at least two cameras, based on the principle that different images are formed in the left eye and the right eye when a person's two eyes are about 6.5 cm apart and look at the subject. According to one embodiment, at least one processor (370) can acquire the depth and size of a subject based on first sensing data including the left eye image and the right eye image. However, it is not limited thereto, and the first sensor (320-1) may be implemented as a Lidar sensor, a Radar sensor, an ultrasonic sensor, an infrared sensor, etc.

[0068] According to one embodiment, the second sensor (320-2) may include a camera including an RGB camera sensor. For example, the camera may convert an image of a subject into an electrical signal and generate image data based on the converted signal. For example, the camera may convert the subject into an electrical image signal through a charge coupled device (CCD), amplify the converted image signal, convert it into a digital signal, and then process the signal. For example, the second sensor (320-2) may include at least one of a standard camera, an RGB camera, and an ultra-wide-angle camera. However, it is not limited thereto, and the second sensor (320-2) may be implemented as a LiDAR sensor, a radar sensor, an ultrasonic sensor, an infrared sensor, etc. According to one embodiment, at least one processor (370) may identify a customer, a store manager, a product, a price tag, an event banner, a stockout indicator, an aisle congestion status, or an employee placement status based on the second sensing data obtained through the second sensor (320-2). According to one embodiment, at least one processor (370) can determine display density, customer dwell density, whether a queue is formed, whether passage is possible, whether an event banner is displayed, or the degree of congestion in a specific area based on second sensing data obtained through a second sensor (320-2).

[0069] According to one embodiment, the first sensor (320-1) may be provided on the main body (330) and may be located on the upper side of the main body (330) to properly identify a customer, store manager, display stand, or obstacle located within the field of view of the first sensor (320-1). According to one embodiment, the second sensor (320-2) may be located on the gripper (311) or manipulator (310) to properly identify a screen, guidance object, document presented by the store manager, or a specific display stand area located adjacent to the gripper (311). However, this is an example for convenience of explanation, and the first sensor (320-1) and the second sensor (320-2) may each be located in various positions on the main body (330).

[0070] According to one embodiment, the drive unit (340) can control the movement of the service robot (300). For example, the drive unit (340) can move the service robot (300), stop the service robot (300) while it is moving, and control the movement speed and direction of movement of the service robot (300). According to one embodiment, the driving type of the service robot (300) may be a wheel type or a walking type. According to one embodiment, the wheel type may refer to a method in which the service robot (300) moves through the rotation of a wheel. If the service robot (300) is a wheel type robot, the service robot (300) may include one or more wheels. The drive unit (340) may include a device that generates power to rotate the wheel. For example, the drive unit (340) may be implemented as an electric motor, a reduction gear, or an actuator. According to one embodiment, the walking type may refer to a method in which the service robot (300) moves through the movement of its legs. If the service robot (300) is of the walking type, the service robot (300) may include two or more legs that support the service robot (300). The legs may include a plurality of links and joints connected to the links. The driving unit (340) may include a device that generates power to lift or lower the legs by rotating the links around the joints. For example, the driving unit (340) may be implemented with a motor and an actuator.

[0071] According to one embodiment, the driving unit (340) can control the movement of a part of the service robot (300). The driving unit (340) may be coupled between a first part of the service robot (300), e.g., the body, and a second part, e.g., the head, arm, or display support structure. The driving unit (340) can rotate the second part. For example, the driving unit (340) can adjust the gaze of the service robot (300) toward the store manager when conversing with the store manager, rotate the display (350) toward the customer when responding to a customer, and adjust the direction of the main body (330) or manipulator (310) when guiding to a specific store location.

[0072] According to one embodiment, the display (350) can be implemented as various types of displays such as an LCD (liquid crystal display), an OLED (organic light-emitting diode), an LCoS (Liquid Crystal on Silicon), a DLP (Digital Light Processing), a QD (quantum dot) display panel, a QLED (quantum dot light-emitting diodes), a μLED (Micro light-emitting diodes), a Mini LED, etc. According to one embodiment, the display (350) may be implemented as a touch screen combined with a touch sensor, a flexible display, a rollable display, a 3D display, a display in which a plurality of display modules are physically connected, etc. According to one embodiment, the display (350) can display status information or service information of a service robot (300) under the control of at least one processor (370). For example, the display (350) may display a symbol or icon representing a guidance action performed by the service robot (300), and may display the remaining battery level of the service robot (300), an error message, an emotional expression, or an animation effect. However, it is not limited thereto, and the display (350) may display product location, discount information, out-of-stock information, event information, congestion information, detour route information, customer-tailored response, store manager input screen, or confirmation request screen, etc., under the control of at least one processor (370). For example, when user input is received by touching the display (350) through an input device such as a stylus pen or one or more fingers, at least one processor (370) may control the service robot (300) according to the user input.

[0073] According to one embodiment, the service robot (300) may include a lift (360). The lift (360) may include a plate, a display board, a guide terminal, or a loading surface on which a guide object may be placed, and may raise or lower the plate under the control of at least one processor (370). According to one embodiment, the lift (360) may adjust the height of the display (350) or the guide terminal to match the eye level of a customer or store manager. According to one embodiment, the lift (360) may temporarily place small guide materials, flyers, sample products, verification tags, or small loading materials. According to one embodiment, at least one processor (370) may raise the lift (360) to make the display screen more visible to the customer and lower the lift (360) to increase stability during movement.

[0074] According to one embodiment, at least one processor (370) can control the operation of the service robot (300) overall. For example, at least one processor (370) can control the operation of the service robot (300) overall by being connected to each component of the service robot (300). According to one embodiment, at least one processor (370) can perform the operation of the service robot (300) according to various embodiments by executing at least one instruction stored in memory (380). According to one embodiment, at least one processor (370) may include one or more of a CPU (Central Processing Unit), GPU (Graphics Processing Unit), APU (Accelerated Processing Unit), MIC (Many Integrated Core), DSP (Digital Signal Processor), NPU (Neural Processing Unit), hardware accelerator, or machine learning accelerator. At least one processor (370) can control one or any combination of other components of the electronic device and can perform operations or data processing related to communication. At least one processor (370) can execute one or more programs or instructions stored in memory (380). For example, at least one processor (370) can perform a method according to one embodiment of the present disclosure by executing one or more instructions stored in memory (380).

[0075] When a method according to one embodiment of the present disclosure includes a plurality of operations, the plurality of operations may be performed by a single processor or by a plurality of processors. For example, when a first operation, a second operation, and a third operation are performed by a method according to one embodiment, the first operation, the second operation, and the third operation may all be performed by a first processor, or the first operation and the second operation may be performed by a first processor, e.g., a general-purpose processor, and the third operation may be performed by a second processor, e.g., an artificial intelligence dedicated processor.

[0076] According to one embodiment, at least one processor (370) may be implemented as a single-core processor including one core, or as one or more multi-core processors including multiple cores. When at least one processor (370) is implemented as a multi-core processor, each of the multiple cores included in the multi-core processor may include internal processor memory such as cache memory or on-chip memory, and a common cache shared by multiple cores may be included in the multi-core processor. Additionally, each of the multiple cores included in the multi-core processor or some of the multiple cores may independently read and execute program instructions for implementing the method according to one embodiment of the present disclosure, or all or some of the multiple cores may be linked to read and execute program instructions for implementing the method according to one embodiment of the present disclosure. When the method according to one embodiment of the present disclosure includes multiple operations, the multiple operations may be performed by one core among the multiple cores included in the multi-core processor, or may be performed by multiple cores. For example, when the first operation, the second operation, and the third operation are performed by a method according to one embodiment, the first operation, the second operation, and the third operation may all be performed by a first core included in a multi-core processor, or the first operation and the second operation may be performed by the first core and the third operation may be performed by the second core.In the embodiments of the present disclosure, a processor may mean a system-on-chip (SoC) in which at least one processor and other electronic components are integrated, a single-core processor, a multi-core processor, or a core included in a single-core processor or a multi-core processor, wherein the core may be implemented as a CPU, GPU, APU, MIC, DSP, NPU, hardware accelerator, or machine learning accelerator, but the embodiments of the present disclosure are not limited thereto.

[0077] According to one embodiment, at least one processor (370) can control the components of the service robot (300). At least one processor (370) can control the components of the service robot (300) by providing signals to each of them, for example, a gripper (311), a sensor (320), a drive unit (340), a display (350), and a lift (360). According to one embodiment, at least one processor (370) can detect or identify customers, store managers, products, display stands, event banners, aisle conditions, or congestion conditions based on sensing data received from the sensor (320), for example, a first sensor (320-1) and a second sensor (320-2). According to one embodiment, at least one processor (370) can collect store information through conversation with the store manager. For example, at least one processor (370) can query the store manager about changes to today's events, changes to discount availability, products that are out of stock or scheduled for restocking, congested areas, areas requiring detours, or matters to be prioritized when guiding customers. According to one embodiment, at least one processor (370) can generate store information based on sensing data and store manager responses. According to one embodiment, at least one processor (370) can analyze the generated store information to update store operation data including product location information, event schedule information, discount application information, out-of-stock information, restocking schedule information, congested area information, detour necessary information, customer precautions information, store manager announcement information, and guidance priority information.

[0078] According to one embodiment, at least one processor (370) can convert updated store operation data into a RAG-based service policy. For example, at least one processor (370) can generate a RAG-based service policy by associating the store operation data with document fragments by search unit, query response keywords, assigned area tags, time tags, strategy type tags, query type tags, and response template identifiers. According to one embodiment, at least one processor (370) can upload the generated RAG-based service policy to a main server or an external server. According to one embodiment, at least one processor (370) can access the RAG-based service policy reflected on the main server and determine a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy. For example, when a customer query is received, at least one processor (370) can generate a response including discount information or congestion information along with location information by referring to the RAG-based service policy. For example, at least one processor (370) can change the circulating route of the assigned area if the congestion level of a specific area increases. For example, at least one processor (370) may re-search the shelf or re-ask the store manager if the reliability of the out-of-stock information is low. For example, at least one processor (370) may receive a shared RAG-based service policy collected by another service robot and perform adjacent area support or customer handover.

[0079] According to one embodiment, the memory (380) may store various information or data associated with the operation of the service robot (300). For example, the memory (380) may include one or more storage media that store at least one instruction. For example, the memory (380) may include instructions that cause the service robot (300) to perform at least one operation when executed individually or collectively by at least one processor (370). According to one embodiment, the memory (380) may store customer history information, store manager response information, sensing data, store information, store operation data, RAG-based service policy, policy version information, reliability score, service strategy, assigned area information, and map information.

[0080] According to one embodiment, the communication unit (390) may include a communication circuit and may support wired or wireless communication between the service robot (300) and an external electronic device, such as a user device or a main server. Wireless communication methods may include, for example, LTE, 5G NR, Wi-Fi, Bluetooth, Bluetooth Low Energy, infrared communication (IrDA), Ultra WideBand (UWB), and Near Field Communication (NFC). According to one embodiment, the communication unit (390) may support communication between a plurality of service robots. According to one embodiment, the communication unit (390) may transmit store information collected by the service robot (300) or a RAG-based service policy generated to the main server, and may receive a RAG-based service policy reflected in the main server or a cooperation request from another service robot.

[0081] According to one embodiment, the service robot (300) may include additional configurations in addition to the illustrated configuration, such as a LiDAR sensor used for driving, a voice input / output device, a turn signal, a light for detecting customer approach, a battery management device, etc., or some configurations may be omitted.

[0082] Figure 4 is a flowchart illustrating the detailed steps of the multi-robot collaboration method of the present invention.

[0083] Referring to FIG. 4, the multi-robot collaboration method may include the step of at least one service robot (300) circulating a preset assigned area within a store (operation 410); the step of the at least one service robot (300) collecting store information by exploring the assigned area (operation 420); the step of the at least one service robot (300) updating store operation data corresponding to the store information by analyzing the store information (operation 430); the step of the at least one service robot (300) converting the store operation data into a RAG-based service policy (operation 440); the step of the at least one service robot (300) uploading the RAG-based service policy to a main server (operation 450); and the step of the at least one service robot (300) accessing the RAG-based service policy and determining a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy (operation 460).

[0084] According to one example, in operation 410, the service robot (300) can circulate a preset assigned area within the store.

[0085] Here, the assigned area may be a spatial range corresponding to at least a part of the entire store area. The assigned area may not be merely a section with an evenly divided area, but may be an operational unit that reflects at least one of changes in product display conditions, customer dwell frequency, potential for congestion, manager accessibility, and potential for concentrated customer inquiries. The assigned area may include at least one of a specific brand zone, a specific display area, an event area, an area adjacent to the checkout counter, an area adjacent to the entrance, a customer waiting area, and a manager stationed area.

[0086] The patrol may be a movement motion for the service robot (300) to repeatedly check the status of the area it is in charge of. The patrol may not be a simple driving motion, but may be an information-gathering movement motion to secure observation opportunities necessary for collecting subsequent store information. The route information may be information including at least one of the outer boundary of the area it is in charge of, an internal passable section, the location of major display shelves, a customer passageway, a location where a manager can respond, and a location requiring re-inspection. The movement path may be a driving path set based on the route information.

[0087] The patrol schedule may be reference information for determining the order and timing of visits to the assigned area based on the time of day or operational conditions. The movement priority may be reference information for determining which of the multiple points within the assigned area will be visited first. The stopping point may be a location where the service robot (300) temporarily stops to check the surrounding conditions more precisely or to increase the possibility of a response from the store manager. The re-search cycle may be a time interval for revisiting the same location or the same area to check for changes. The avoidance path may be an alternative path to maintain patrols of the assigned area in exceptional situations such as customer congestion, aisle blockage, the occurrence of obstacles, or the placement of temporary materials.

[0088] The service robot (300) can generate a movement path based on path information of the area it is responsible for. When generating the movement path, the service robot (300) can adjust the movement path so that it includes points with high importance for information collection, rather than just reflecting the spatial structure of the area it is responsible for. For example, if there is a display stand where the display status changes frequently, the service robot (300) can set the movement path to pass through that display stand more frequently. For example, if there is a location where customer inquiries frequently occur, the service robot (300) can set the movement path to include that location.

[0089] The service robot (300) can repeatedly circulate its assigned area by reflecting a circulating schedule, movement priority, stopping point, re-search cycle, and avoidance path. For example, the service robot (300) can set a circulating schedule to prioritize checking the event preparation status, display status, and discount application status immediately after opening. For example, the service robot (300) can set a circulating schedule to prioritize checking customer dwell density, whether queues are formed, and whether passage is possible during times when congestion is likely. For example, the service robot (300) can set a circulating schedule to prioritize checking out-of-stock status, stock shortage status, and operator announcements during times adjacent to closing.

[0090] The service robot (300) may apply different circulation orders even within the same assigned area according to movement priority. For example, the service robot (300) may set the movement priority of the event area high if there is a high possibility of a change in the state of the event area. For example, the service robot (300) may set the movement priority of the customer inquiry concentration area high to increase the possibility of responding to the customer inquiry concentration area.

[0091] The service robot (300) can obtain visual information, voice information, object information, and location information with higher precision at the stopping point. The stopping point may correspond to at least one of the following: a position observing the front of a display stand, a location near a manager, a location near event information, or a location with a high frequency of customer stay. The service robot (300) can determine whether it is possible to converse with the store manager at the stopping point and can prepare the timing for the conversation required during the store information collection stage.

[0092] The service robot (300) can perform continuous status checks rather than one-time checks by repeatedly visiting the same point according to the re-search cycle. The re-search cycle can be set differently depending on the characteristics of the area in charge. Event areas, areas adjacent to checkout counters, and areas adjacent to entrances may have short re-search cycles as they may be areas where the state changes quickly. General display areas and brand zones with relatively little change may have relatively long re-search cycles.

[0093] The service robot (300) can select an avoidance path when it detects a customer cluster or an obstacle. The avoidance path can be set so that gaps in information collection are minimized even in exceptional situations that occur during the patrol of the assigned area. Even after selecting the avoidance path, the service robot (300) can adjust the timing of re-access or the order of re-inspection so that key information collection points are not missed. The service robot (300) can correct the path so that the continuity of information collection in the assigned area is maintained, rather than changing the path based solely on movement efficiency.

[0094] Multiple service robots (300) can circulate in parallel by dividing the entire area of ​​the store. Multiple service robots (300) can each correspond to different assigned areas. For example, a first service robot can circulate an assigned area corresponding to an entrance or an event area, a second service robot can circulate a specific brand zone or a display area with frequent inventory changes, and a third service robot can circulate an area adjacent to the checkout counter or a customer waiting area. In this way, by performing distributed circulation, the inspection cycle for the entire store can be shortened.

[0095] Multiple service robots (300) can coordinate their movement by considering their mutual positional relationships within the boundaries of their respective areas. For example, the multiple service robots (300) can reduce the possibility of collision by adjusting at least one of speed, stopping time, and direction of movement near the boundaries of their respective areas. For example, the multiple service robots (300) can increase the efficiency of searching the entire store by reducing redundant circulating near the boundaries of their respective areas.

[0096] Multiple service robots (300) can variably adjust the division of assigned areas according to exceptional circumstances. For example, adjacent service robots may temporarily supplement the assigned area if a specific service robot stays for a long time attending to customers, is low on battery, or is focused on a conversation with a manager. For example, by performing both the division of assigned areas and supplementary patrols, multiple service robots (300) can reduce gaps in information collection for the entire store.

[0097] According to one example, in operation 420, the service robot (300) can collect store information by exploring the area in charge.

[0098] Here, store information may be information indicating the current operational status of the assigned area. Area exploration may be an act in which the service robot (300) moves within the assigned area to identify objects of observation and objects to respond to. Interactive store information may be information collected based on question-and-answer with a store manager or on-site operator. Non-interactive store information may be information collected based on sensor detection, image analysis, object recognition, and location recognition. Composite store information may be information combined with interactive store information and non-interactive store information. The reliability score may be a value that numerically indicates the reflectability of the composite store information.

[0099] The service robot (300) can collect store information while moving along key observation locations, manager response locations, customer dwelling locations, and display status check locations within the assigned area. The service robot (300) can more precisely check the status of a specific location by performing at least one of deceleration, stopping, changing direction, and re-inspection during the search process. The service robot (300) can prioritize securing information having operational significance that can be used for subsequent operational data updates during the circuitous route of the assigned area.

[0100] FIG. 5 is a diagram illustrating the operation of the service robot (300) of the present invention collecting store information.

[0101] In one embodiment, the service robot (300) can collect conversational store information by actively querying the store manager (10), as illustrated in FIG. 5. The store manager (10) may provide at least one of event information, discount information, out-of-stock information, restocking schedule information, congestion information, detour information, and customer guidance priority information. The service robot (300) may convert the utterance of the store manager (10) into text using a voice recognition module, and the service robot (300) may extract at least one of an event name, product name, brand name, discount rate, application time, location expression, congestion expression, and detour need expression from the text conversion result. The service robot (300) may not perform only a single question, but may add a follow-up question based on the content of the preceding response. When a response regarding a specific brand event is collected, the service robot (300) may query again for at least one of a discount rate, application start time, application end time, the location of the corresponding brand, and a customer notification phrase. If the response content is unclear, the service robot (300) can make a re-inquiry to the same manager (10) or another manager.

[0102] In one embodiment, the service robot (300) can collect non-interactive store information by performing sensor-based detection. For example, the service robot (300) can acquire the visual status of the area it is responsible for using a camera module. For example, the service robot (300) can check the passability status and space occupancy status using a distance sensor module and a depth sensor module. For example, the service robot (300) can identify at least one of products, display stands, banners, customer clusters, queues, aisle obstacles, and staff locations using an object recognition module. By detecting the display stand status, the service robot (300) can check at least one of the display stand front alignment status, whether empty spaces have occurred, whether there is a shortage of products at a specific location, and whether price tags or guide signs are present. By detecting the product display density, the service robot (300) can determine the possibility of being out of stock or the possibility of needing to replenish. By detecting whether an event banner is exposed, the service robot (300) can determine whether a specific event is actually taking place. The service robot (300) can identify areas of interest or potential congestion by detecting customer dwell density. The service robot (300) can identify potential delays in customer service or movement restrictions by detecting whether a queue is forming. The service robot (300) can identify areas requiring detours by detecting whether passage is possible. Additionally, the service robot (300) can determine whether it is appropriate to hand over customers to an employee within the assigned area or whether the store manager can handle the situation by detecting the employee placement status.

[0103] In one embodiment, the service robot (300) can generate composite store information for the same assigned area by combining conversational store information and non-conversational store information. For example, the service robot (300) can generate composite store information with a high probability of being out of stock by combining the store manager's out-of-stock response and the shelf empty detection result. For example, the service robot (300) can generate composite store information regarding the congestion state by combining the store manager's congestion response and the customer dwell density detection result. The service robot (300) can determine the reliability of the composite store information to be high if the conversational store information and the non-conversational store information match each other. If the conversational store information and the non-conversational store information do not match each other, the service robot (300) can classify the store information as requiring reconfirmation.

[0104] The reliability score C can be calculated according to [Formula 1] below.

[0105] [Formula 1]

[0106] C = w1*M + w2*V + w3*X + w4*R

[0107] Here, parameter M may be a value indicating whether the store manager has directly verified the information. Parameter V may be a value indicating whether the interactive collection results and the non-interactive collection results match. Parameter X may be a value indicating whether cross-verification occurs between multiple service robots. Parameter R may be a value indicating a freshness score based on the collection time. Weights w1, w2, w3, and w4 may be values ​​for adjusting the weight of each parameter. The service robot (300) may use complex store information with a high reliability score more preferentially for subsequent operational decisions.

[0108] In one embodiment, the service robot (300) may set the magnitudes of weights w1, w2, w3, and w4 differently depending on the type of store information or the purpose of utilizing the store information. For example, the service robot (300) may set the weight w1, corresponding to parameter M, to the highest value for event information, discount information, out-of-stock information, and restocking schedule information, where direct delivery by the store manager is critical. For example, the service robot (300) may set the weight w2, corresponding to parameter V, to the highest value for event banner exposure status, product display status, and imminent out-of-stock status, where mutual consistency between conversational collection results and non-conversational collection results is important. For example, the service robot (300) may set the weight w3, corresponding to parameter X, to the highest value for congestion information, detour information, and passage availability status, which are independently verified by multiple service robots (300) near the boundaries of adjacent or same service areas. For example, the service robot (300) may set the weight w4 corresponding to parameter R to the highest value for customer dwell density, whether a queue is formed, information on congested areas, and information requiring detours, which can fluctuate rapidly over time. For example, the service robot (300) may preferentially set the weight w1 or weight w4 to a high value for information requiring high customer safety or immediate guidance. For example, the service robot (300) may preferentially set the weight w3 to a high value for information where consistency between multiple service robots (300) is important. For example, the service robot (300) may calculate a reliability score C in a manner optimized for the nature of the information, even for the same complex store information, based on the above-mentioned weight adjustments.

[0109] Multiple service robots (300) can independently collect store information in different assigned areas. For example, a first service robot can collect event information and customer density information in an event area. For example, a second service robot can collect discount information and out-of-stock information in a specific brand zone. For example, a third service robot can collect information on whether a queue is formed and the need for a detour in an area adjacent to the checkout counter. Multiple service robots (300) can use the store information obtained from their distributed assigned areas as basic information for forming a common policy.

[0110] According to one example, in operation 430, the service robot (300) can update store operation data corresponding to the store information by analyzing the store information.

[0111] Here, store operation data may be operational standard information structured to allow direct use of store information for customer guidance, determining service availability, deciding service delivery methods, and forming subsequent service policies. Operation information may be item-specific information extracted from store information. Product location information may be information indicating the location where products or brands are placed. Event schedule information may be information indicating the event start time, event end time, and applicable areas. Discount application information may be information indicating the discount rate, eligible items, and discount period. Out-of-stock information may be information indicating the current out-of-stock status of a product. Restock schedule information may be information indicating the restock time or expected restock status. Congested area information may be information indicating areas where customer density or queue formation has occurred. Detour required information may be information regarding areas where traffic restrictions or movement obstacles exist. Customer notice information may be information indicating precautions that need to be prioritized for customer guidance. Managerial message information may be information indicating operational changes separately communicated by the store manager. Guidance priority information may be information indicating the item that must be reflected first among multiple guidance items. Update conditions may serve as criteria for determining whether to reflect new operational information in existing store operation data. Text conversion of speech recognition data may be a process for converting a store manager's utterances into text form. Classification of object recognition data may be a process for distinguishing products, display shelves, banners, customer clusters, queues, obstacles, and employee locations. Mapping with location information may be a process for connecting extracted information to specific locations within a designated area. Event occurrence determination may be a process for determining whether collected information signifies a change in actual store operations.Time information linkage may be a process for connecting the collection time, application start time, application end time, and last verification time. Area identification may be a process for specifying the spatial area within the scope of information application. Priority determination may be a process for determining which items among multiple pieces of operational information should be reflected first.

[0112] The service robot (300) can update store operation data by analyzing the previously collected conversational store information, non-conversational store information, and complex store information. The service robot (300) can convert the store manager's speech into text using a voice recognition module, and the service robot (300) can extract at least one of a product name, brand name, discount rate, event start time, event end time, out of stock status, restocking scheduled time, detour required section, and customer guidance priority phrase from the text conversion result. The service robot (300) can classify at least one of a product, display stand, banner, customer cluster, queue, aisle obstacle, and employee location using an object recognition module. The service robot (300) can map which location within which area the extracted information corresponds to using a location recognition module. The service robot (300) can assign at least one of a collection time, an application start time, an application end time, and a last check time to each piece of information by performing time information linkage.

[0113] The service robot (300) can determine whether the collected information is an actual operational change by performing an event occurrence determination. For example, if the service robot (300) includes "30 percent sale starting today" in the store manager's utterance, it can interpret the information as a new occurrence of event schedule information and discount application information. For example, if the service robot (300) confirms whether the event banner is displayed, it can determine the certainty of the event occurrence to be higher. For example, if the service robot (300) confirms an increase in customer dwell density and the formation of a queue together, it can determine that information regarding a congested area has occurred. For example, if the service robot (300) confirms a decrease in the passability of a specific passageway and a detour guidance response together, it can determine that information regarding the need for a detour has occurred.

[0114] The service robot (300) can specify the spatial coverage of each operational information by performing area identification. For example, the service robot (300) can link discount information collected in a specific brand zone with area information corresponding to that brand zone. For example, the service robot (300) can specify congestion information collected in an area adjacent to the checkout counter as congestion zone information of the area adjacent to the checkout counter. For example, the service robot (300) can give higher priority to congestion zone information or information requiring detours that directly affect customer safety and movement paths by performing priority determination. For example, the service robot (300) can give high guidance priority to event information or discount information that is expected to have a high frequency of customer inquiries.

[0115] The service robot (300) can extract operational information based on the analysis results. For example, the service robot (300) can extract at least one of product location information, event schedule information, discount application information, out-of-stock information, restocking schedule information, congested area information, detour required information, customer precautions information, manager delivery information, and guidance priority information from store information. For example, if the service robot (300) confirms that conversational store information such as "there are almost no specific products" and an increase in empty spaces on the corresponding display stand are confirmed together, it can extract out-of-stock information or a stock shortage status. For example, if the service robot (300) confirms that conversational store information such as "a specific brand sale starting this afternoon" and the display of an event banner are confirmed together, it can extract event schedule information and discount application information. For example, if the service robot (300) confirms that customer crowding in front of the checkout counter and a response requiring detour guidance are confirmed together, it can extract congested area information and detour required information.

[0116] In one embodiment, the service robot (300) can compare the extracted operational information with previously stored store operational data. The service robot (300) can add new items for items that did not previously exist. The service robot (300) can modify existing items if the content of the existing items has changed. The service robot (300) can remove existing items if the application period of the existing information has expired or if the existing information is no longer valid. The service robot (300) can maintain the content of the existing information but adjust only the priority. The service robot (300) can update the application period if the event end time is extended.

[0117] In one embodiment, the service robot (300) may update store operation data only when pre-set update conditions are met. For example, the update conditions may include at least one of the following: whether a direct response from a store manager exists, whether conversational store information and non-conversational store information match, whether data is collected independently among multiple service robots, whether the validity period of existing data has expired, and whether the reliability of new information is superior. The service robot (300) may retain existing data if information formed solely by a single non-conversational detection exists. The service robot (300) may reflect new operation information if both the manager's response and the detection result exist, and cross-confirmation by another service robot also exists. The service robot (300) may replace existing data with new information if the reliability score of the new information is sufficiently higher than the reliability score of the existing data. If the difference in reliability of the new information is insufficient, the service robot (300) may retain existing data and wait for reconfirmation.

[0118] The service robot (300) can also form a connection between multiple operational information during the process of updating store operation data. The service robot (300) can store discount application information in connection with product location information and event schedule information. The service robot (300) can store congested area information in connection with information on detours and customer precautions information. The service robot (300) can store manager delivery information in connection with guidance priority information.

[0119] According to one example, in operation 440, the service robot (300) can convert the store operation data into a RAG-based service policy.

[0120] Here, RAG-based service policies may be policy information including searchable knowledge information for determining customer queries or services provided, application condition information serving as criteria for judging whether or how a service is provided, and service strategies defining execution directions for each service strategy. Store operation data may be operational standard information used for creating subsequent policies, such as product location information, event schedule information, discount application information, out-of-stock information, restocking schedule information, congestion area information, detour requirements information, customer precautions information, manager messages information, and guidance priority information. Entity information may be information representing the central target of queries and judgments, such as product name, brand name, event name, and zone name. Responsible zone information may be information representing the spatial range to which operational information is applied first. Time information may be information representing the valid range of operational information, such as the application start time, application end time, update time, and verification time. Query type information may be information representing the type of customer query, such as location query, discount query, out-of-stock query, congestion query, and detour query. Strategy type information may be information indicating the type of service strategy to which the policy will be used, such as for circuit strategy, response strategy, interaction strategy, operation strategy, or cooperation strategy.

[0121] In one embodiment, the service robot (300) can convert the store operation data into a RAG-based service policy. The service robot (300) can reconstruct the store operation data into a policy form that can be directly used for searching, judgment, and execution, rather than storing the data as is. During the conversion process, the service robot (300) can classify the store operation data into entity information, assigned area information, time information, query type information, and strategy type information. Using the classification results, the service robot (300) can distinguish between information directly used for customer response, information determining whether to provide a service, and information used for service execution.

[0122] In one embodiment, the service robot (300) can generate knowledge information by extracting information that is directly used for customer responses. The knowledge information may be factual information that can be directly provided to the customer. The service robot (300) may configure at least one of the following as knowledge information: location information of a specific brand, discount rate of a specific product, whether a specific event is in progress, the out-of-stock status of a specific product, and the congestion status of a specific passageway. The service robot (300) may configure searchable location guidance knowledge information by linking operational information regarding the brand location with a location query type. The service robot (300) may configure discount guidance knowledge information by linking discount information and event schedule information with a discount query type. The service robot (300) may configure out-of-stock guidance knowledge information by linking out-of-stock information and restocking schedule information with an out-of-stock query type.

[0123] In one embodiment, the service robot (300) may generate application condition information by extracting information that serves as a criterion for determining whether to provide a service or the method of providing the service. The application condition information may be information that determines when, in which assigned area, and under what circumstances the knowledge information should be used. The service robot (300) may set a time condition of "before the end time of the event" for specific discount information. The service robot (300) may set a spatial condition of "entering a specific assigned area" for specific congestion information. The service robot (300) may set a validity condition of "before the scheduled time of restocking" for specific out-of-stock information. The service robot (300) may set an application condition of "specific time zone" or "specific customer type" for specific customer precautions.

[0124] In one embodiment, the service robot (300) may generate a service strategy by extracting information used for service execution. The service strategy may be information that defines the direction of the service robot (300)'s actual movement, response, interaction, reconfirmation, and collaboration actions. If the customer dwell density in the event area is high and the priority of event information is high, the service robot (300) may generate a service strategy for a circulating strategy and use it as a criterion for increasing the frequency of re-inspection. If a passageway congestion status and information requiring a detour exist together, the service robot (300) may generate a service strategy for a response strategy or a circulating strategy and use it as a criterion for guiding customers to detour or adjusting movement paths. If the importance of out-of-stock information is high and reliability reinforcement is required, the service robot (300) may generate a service strategy for a work strategy and use it as a criterion for re-search or manager reconsideration. If an adjacent area congestion status and the response status of another service robot exist together, the service robot (300) may generate a service strategy for a collaboration strategy and use it as a criterion for requesting support in an adjacent area or customer handover.

[0125] The service robot (300) can generate multiple policy elements in parallel from a single store operation data. The service robot (300) can generate, respectively, knowledge information for customer response, application condition information indicating event validity conditions, and service strategies indicating event area re-inspection criteria from specific brand discount information. The service robot (300) can generate, respectively, congestion guidance knowledge information, detour application condition information, and service strategies serving as criteria for circulating adjustment or cooperative adjustment from specific congestion area information. The service robot (300) can enable various strategy judgments in subsequent stages by dividing the same operation information by purpose and reorganizing it into a policy structure.

[0126] The service robot (300) can establish a connection relationship between query type information and strategy type information. The service robot (300) can set location query types in conjunction with product location information. The service robot (300) can set discount query types in conjunction with discount application information and event schedule information. The service robot (300) can link out-of-stock query types with out-of-stock information and restocking schedule information. The service robot (300) can set congestion query types in conjunction with congestion zone information. The service robot (300) can link detour query types with detour necessary information. The service robot (300) can set strategy types for touring strategies in conjunction with congestion zone information, event concentration zone information, and re-inspection necessary zone information. The service robot (300) can set strategy types for response strategies in conjunction with response template information. The service robot (300) can set strategy types for response strategies in conjunction with whether to provide preemptive guidance or response priority criteria. The service robot (300) can set the strategy type for work strategy in conjunction with re-search, re-verification, and material requirement criteria. The service robot (300) can set the strategy type for cooperation strategy in conjunction with role sharing, support request, and customer handover criteria.

[0127] A response template identifier may be information that identifies the sentence format or output format used when a retrieved policy is reflected in a customer response. A service robot (300) can match the response template identifier with each piece of knowledge information. The service robot (300) can output a discount guidance sentence by assigning a response template identifier corresponding to discount information. The service robot (300) can output a detour guidance sentence by assigning a response template identifier corresponding to congestion information. The service robot (300) can output a sentence that provides information on the fact of being out of stock and the time of restocking by assigning a response template identifier corresponding to out-of-stock information and restocking schedule information.

[0128] The service robot (300) can configure generated knowledge information, application condition information, and service strategy into a searchable form. The service robot (300) can set entity information, query type information, assigned area information, and time information as search keys. The service robot (300) can assign priority scores and reliability scores to each policy element. The service robot (300) can set the policy structure so that policies highly relevant to the current query situation or current operation situation are searched first. The service robot (300) can set the policy structure so that policies highly relevant to recency or assigned area are considered first in subsequent judgments.

[0129] Multiple service robots (300) can convert store operation data collected and organized in different assigned areas into a common policy format. For example, the first service robot can convert discount information collected in the event area into knowledge information, application condition information, and service strategy. For example, the second service robot can convert out-of-stock information collected in a specific brand zone into knowledge information, application condition information, and service strategy. For example, the third service robot can convert congestion information collected in the area adjacent to the checkout counter into knowledge information, application condition information, and service strategy. By using the common policy format, the multiple service robots (300) can utilize operation information formed in different assigned areas in the same way.

[0130] In this way, the service robot (300) can convert store operation data into a directly applicable RAG-based service policy. That is, by reflecting the RAG-based service policy in the main server or RAG storage area, the service robot (300) can enable other service robots to access the RAG-based service policy and apply the latest updated service policy in real time.

[0131] According to one example, in operation 450, the service robot (300) can upload the RAG-based service policy to the main server.

[0132] Here, the main server may be a server device that receives and stores RAG-based service policies transmitted from multiple service robots (300). The assigned area identification information may be information indicating the spatial range to which the RAG-based service policy corresponds preferentially. The collection time information may be information indicating at what point in time the field status reflected by the RAG-based service policy was collected. The service robot identification information may be information indicating the identification information of the service robot (300) that created the RAG-based service policy. The policy version information may be information for distinguishing the order of policy updates when the same policy item is repeatedly updated. The update reason information may be information indicating the cause for which the RAG-based service policy was created or modified. The reliability information may be information for determining the reflection strength or the degree of priority utilization of the RAG-based service policy.

[0133] FIG. 6 is a diagram illustrating the operation of the service robot (300) of the present invention uploading an updated RAG-based service policy.

[0134] Referring to FIG. 6, the service robot (300) can transmit the generated RAG-based service policy to the main server (140) rather than storing it only in internal memory. During the process of uploading the RAG-based service policy, the service robot (300) can transmit not only the policy content but also additional information indicating the source and scope of application of the policy. Through the transmission process, the service robot (300) can convert field information obtained from the assigned area into a common policy asset that can be jointly utilized by multiple service robots (300).

[0135] In one embodiment, the service robot (300) may upload a RAG-based service policy based on at least one of a customer urgent request, identification of a stockout status, exceeding a standard congestion level, a preset period, or a store manager command. A customer urgent request may be a situation where it is determined that the latest policy, which must be immediately reflected in a customer query, is required. Identifying a stockout status may be a situation where a shortage of stock or a stockout status of a specific product is confirmed and must be quickly reflected in the response of another service robot (300). Exceeding a standard congestion level may be a situation where the customer dwell density or the degree of queue formation exceeds a preset standard, requiring adjustment of movement or guidance methods. A preset period may be a standard for organizing and synchronizing accumulated policies at regular time intervals. A store manager command may be an operational command requiring immediate sharing of the policy, such as the start of a specific event, prioritizing specific guidance, or concentrating response in a specific area.

[0136] In one embodiment, the service robot (300) can upload policies with high urgency and general policies differently. The service robot (300) can upload policies with high immediacy, such as out-of-stock information, congestion information, and urgent notice information, in real time. The service robot (300) can upload policies with relatively low immediacy, such as general display change information or periodic operation information, at regular intervals. By distinguishing the upload methods, the service robot (300) can secure both real-time performance and communication efficiency.

[0137] In one embodiment, the service robot (300) can organize RAG-based service policies into uploadable units. The service robot (300) can organize a policy structure including knowledge information, application condition information, and service strategy into one or more policy items. The service robot (300) can link a responsible area identification information, collection time information, service robot identification information, policy version information, update reason information, and reliability information to each of the policy items. The service robot (300) can generate an upload record including the link information.

[0138] The service robot (300) can link the area identification information of the brand zone in charge of the discount policy generated in a specific brand zone. The area identification information in charge can be used as a criterion to distinguish the spatial coverage of the policy. The service robot (300) can link the area identification information of the area adjacent to the checkout counter to the congestion policy generated in the area adjacent to the checkout counter. The main server (140) can distinguish the spatial relevance of the policy using the area identification information in charge.

[0139] The service robot (300) can provide collection time information along with policies that change rapidly over time, such as event information, out-of-stock information, congestion information, and detour information. The collection time information can be used as a criterion for determining the latest status of the policies. The main server (140) can determine which policy reflects the more recent site conditions by referring to the collection time information.

[0140] The service robot (300) can make it possible to identify which service robot (300) created which policy by transmitting the service robot identification information to the main server (140). The service robot identification information can be used as a criterion for distinguishing the policy creator. The main server (140) can track the source of the policy using the service robot identification information. The main server (140) can also manage the cross-verification history between multiple service robots (300) using the service robot identification information.

[0141] The service robot (300) can distinguish the initial policy, the modified policy, and the termination policy into different versions when the same brand event policy is repeatedly modified. The policy version information can indicate the update order of the same policy item. The main server (140) can determine which policy is the more recent updated version using the policy version information.

[0142] The service robot (300) may record at least one of the following as update reason information: detection of a new event, identification of a sold-out status, increased congestion, direct command by an administrator, result of reconfirming inventory, and confirmation of a match between interactive and non-interactive information. The update reason information may indicate the cause of policy creation or modification. The main server (140) may store the policy formation context together by referring to the update reason information.

[0143] A service robot (300) can generate reliability information by reflecting at least one of the following: whether there is a direct response from an administrator, whether conversational information and non-conversational information match, whether there is cross-checking between multiple service robots (300), and recency. The reliability information may serve as a criterion for determining the priority of policy utilization and the strength of reflection. The main server (140) can adjust the degree of priority reflection of the policy by referring to the reliability information.

[0144] The main server (140) can store the uploaded RAG-based service policy in a state where the policy content and policy metadata are combined. During the storage process, the main server (140) can manage the assigned area identification information, collection time information, service robot identification information, policy version information, update reason information, and reliability information together with the RAG-based service policy. The main server (140) can maintain the source, time, version, reason, and reliability of the policy in a traceable manner using the storage structure.

[0145] The main server (140) can manage each policy separately when multiple identical or similar policies are uploaded. The main server (140) can determine which policy reflects a more recent state by referring to collection time information and policy version information. The main server (140) can determine which policy has a more direct application relationship by referring to reliability information and area identification information. Based on the above management method, the main server (140) can maintain policy consistency at the stage where the service robot (300) downloads and utilizes the service policy.

[0146] In one embodiment, the service robot (300) can adjust the upload method according to the communication status. The service robot (300) can perform an upload immediately when the communication connection with the main server (140) is stable. When the communication connection with the main server (140) is unstable, the service robot (300) can temporarily store a RAG-based service policy in a policy buffer area. When the communication connection is restored, the service robot (300) can sequentially upload policy items in the policy buffer area.

[0147] In one embodiment, the service robot (300) can upload only changed items, excluding items that duplicate policies already uploaded. The service robot (300) can determine whether a change has been made by comparing policy version information with update time information. The service robot (300) can reduce unnecessary communication volume by transmitting only changed items. The main server can efficiently accumulate only the latest policies by using the change item upload structure.

[0148] In one embodiment, the main server can reflect the uploaded RAG-based service policy in a common policy repository. The main server can maintain the common policy repository as a policy asset that can be jointly utilized by a plurality of service robots (300) in a subsequent service strategy determination step. Accordingly, the service robots (300) can extend individual service policies generated in their assigned areas into a common service policy for the entire multi-robot community through the main server.

[0149] According to one example, in operation 460, the service robot (300) can access the RAG-based service policy and determine a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy.

[0150] Here, the service strategy may be a criterion for determining at least one of the directions of movement, response, interaction, reconfirmation, and collaboration that the service robot (300) must perform at the current time. The query condition may be a search condition that reflects at least one of the following: the status of the assigned area, congestion level, event status, content of the customer inquiry, whether the customer is accessible, whether there is a request from an administrator, time of day, customer type, and the operational status of other service robots. The priority score may be a value indicating which of the multiple policies found should be applied first in the current situation. The reliability score may be a value indicating how reliable the found policy is based on field information. The circulation strategy may be a strategy for determining at least one of the following: the frequency of re-inspection, the detour area, the movement path, and the priority area for circulation. The response strategy may be a strategy for determining the content of the response to be included in the customer inquiry, the order of the response, and the format of the response. The interaction strategy may be a strategy for determining at least one of the following: whether to provide preemptive guidance, the priority of the response, the level of explanation, and the tone of the response. The work strategy may be a strategy for determining the necessity of at least one of re-search, re-verification, store manager inquiry, event banner verification, and repeated collection. The cooperation strategy may be a strategy for determining at least one of role sharing, requesting information re-verification, requesting support for adjacent areas, customer handover, and temporary joint patrol.

[0151] FIG. 7 is a diagram illustrating the operation of a service robot (300) of the present invention determining a service strategy based on RAG.

[0152] Referring to FIG. 7, the service robot (300) can access RAG-based service policies stored on the main server. The service robot (300) can not only follow its own internal fixed rules but can also search for and apply policies related to the current situation. For example, the service robot (300) can apply not only information directly collected in its assigned area but also policies formed in other areas assigned by other service robots.

[0153] The service robot (300) can configure query conditions that reflect the current state. The service robot (300) can include at least one of the following in the query conditions: current location, current assigned area, current time, current customer inquiry content, current congestion level, current event status, current administrator request status, and current adjacent robot status. The service robot (300) can configure query conditions that include a location query type when a customer asks about a specific brand location. The service robot (300) can configure query conditions that include both a discount query condition and a congestion condition when a customer asks about a discount and the congestion level in the direction of the checkout counter is high. The service robot (300) can configure query conditions that include both a customer access condition and an event condition when customer access is frequent and an event is in progress.

[0154] The service robot (300) can search for RAG-based service policies corresponding to query conditions. The service robot (300) can search for service policies by considering at least one of query type information, assigned area information, time information, strategy type information, entity information, priority score, and reliability score. The service robot (300) can search for a set of related service policies necessary for current service judgment, rather than searching for a single item. When a brand location query is entered, the service robot (300) can search for discount application information, out-of-stock information, congested area information, and detour requirement information together with the brand location information. The service robot (300) can compare the searched set of service policies to select the service policy most suitable for the current situation.

[0155] The service robot (300) can determine the order of service policy application based on priority scores and reliability scores. The service robot (300) can assign high priority to service policies directly related to customer inquiries. The service robot (300) can also assign high priority to congestion policies or detour policies directly related to customer safety or movement restrictions. The service robot (300) can apply service policies with high reliability scores first. If there are service policies with high priority scores but low reliability scores, the service robot (300) can also determine additional work strategies.

[0156] FIG. 8 is a diagram showing an example of a service robot (300) of the present invention determining a circulating strategy based on RAG.

[0157] Referring to FIG. 8, the service robot (300) can determine a patrol strategy. The service robot (300) can search for policies that reflect the status of the assigned area, congestion, event status, and the operation status of other service robots. Based on the priority score and reliability score of the searched policies, the service robot (300) can determine at least one of the re-inspection frequency, bypass area, movement path, and priority area for patrol. If the customer dwell density in the event area is high and the frequency of event-related questions is high, the service robot (300) can increase the re-inspection frequency of the event area. If the area adjacent to the checkout counter is identified as a congested area, the service robot (300) can replace that passageway with an avoidance path. If another adjacent service robot is attending to customers for a long time, the service robot (300) can adjust the movement path to supplement the patrol of a part of the assigned area of ​​the other adjacent service robot.

[0158] FIG. 9 is a diagram showing an example of a service robot (300) of the present invention determining a response strategy based on RAG.

[0159] Referring to FIG. 9, the service robot (300) can determine a response strategy. The service robot (300) can search for a RAG-based service policy that includes a query type tag corresponding to a customer query. The service robot (300) can determine the response content based on at least one of the response template identifier, priority score, product location information, discount application information, out-of-stock information, congestion zone information, and detour need information included in the searched policy. If a customer asks for the final price with a discount applied, the service robot (300) can configure the response content to include the final price with a real-time discount rate applied to the cost price. If a customer asks for the location of Brand A, the service robot (300) can configure the response content to include Brand A's location information, as well as Brand A's discount status, congestion status, and detour need. If a customer asks for the stock of a specific product, the service robot (300) can configure the response content to include out-of-stock information and restocking schedule information. The service robot (300) can determine which information to provide first based on the priority score of the searched policy.

[0160] FIG. 10 is a diagram showing an example of a service robot (300) of the present invention determining a response strategy based on RAG.

[0161] Referring to FIG. 10, the service robot (300) can determine a response strategy. The service robot (300) can search for RAG-based service policies corresponding to customer access, manager requests, time zone, congestion level, and customer type. Based on the searched policies, the service robot (300) can determine at least one of whether to provide preemptive guidance, the priority of response to customers or managers, the level of explanation, and the tone of voice for response. If customer access is frequent in the event area and the priority of event guidance is high, the service robot (300) can start providing event guidance even before the customer asks a question first. If an urgent request from a manager is detected, the service robot (300) can perform response to the manager before responding to general customers. If the service robot (300) detects a type of customer such as the elderly or those requiring detailed guidance, it can increase the level of explanation and apply a friendly and detailed tone of voice for response.

[0162] FIG. 11 is a diagram showing an example of a service robot (300) of the present invention determining a work strategy based on RAG.

[0163] Referring to FIG. 11, the service robot (300) can determine a work strategy. The service robot (300) can search for policies that reflect whether there is an information discrepancy, whether there is a lack of reliability, whether there is a shortage of stock, whether there is a change in the event status, and whether manager verification is required. Based on the searched policies, the service robot (300) can determine the necessity of at least one of re-searching, re-verification, store manager query, event banner verification, and repeated collection. If the non-interactive detection result indicates a possibility of being out of stock and there is no interactive verification, the service robot (300) can re-search the corresponding display stand. If the event banner removal status and the manager response appear different from each other, the service robot (300) can perform event banner verification and store manager re-query together. The service robot (300) can set policies with insufficient reliability scores as targets for additional collection rather than reflecting them in the immediate response.

[0164] FIG. 12 is a diagram showing an example of a service robot (300) of the present invention determining a cooperation strategy based on RAG.

[0165] Referring to FIG. 12, the service robot (300) can determine a cooperation strategy. The service robot (300) can search for policies that reflect the status of other service robots' assigned areas, congestion status, response status, support request information, and customer handover requirements. Based on the searched policies, the service robot (300) can determine at least one of role sharing, information re-verification request, adjacent area support request, customer handover, or temporary joint patrol. The service robot (300) can request support from other service robots. The service robot (300) can accept or reject support requests from other service robots. For example, if the congestion in an adjacent assigned area increases rapidly, the service robot (300) can respond to an adjacent area support request and share some patrol duties. For example, if it is difficult to perform additional verification because another service robot is busy responding to customers, the service robot (300) can accept an information re-verification request and check the inventory status or banner status on its behalf. For example, if the service robot (300) is in charge of a specific customer's inquiry, it can transfer the customer to another service robot.

[0166] In one embodiment, the service robot (300) may determine the final action by combining multiple strategies, rather than independently determining the circulating strategy, response strategy, interaction strategy, work strategy, and cooperation strategy. When the customer density state of the event area, the discount information update state, and the adjacent robot interaction state exist together, the service robot (300) may determine together a circulating strategy that increases the frequency of re-examining the event area, a response strategy including sale information, an interaction strategy that performs preemptive guidance, a work strategy that re-checks the status of the event banner, and a cooperation strategy that performs support for adjacent areas. The service robot (300) may perform actual movement, actual response, actual interaction, actual reconfirmation, and actual collaboration based on the result of combining multiple strategies.

[0167] The service robot (300) can reflect the above judgment results in subsequent field execution. The service robot (300) can perform extended judgments that go beyond the limits of its assigned area based on a common RAG-based service policy. Consequently, the service robot (300) can perform situation-adaptive services based on policy assets formed jointly by multiple service robots.

[0168] As such, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure allows each service robot to repeatedly circulate its assigned area to collect both conversational and non-conversational store information, and to update store operation data by reflecting the reliability of the complex store information. Since the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure may include a structure that considers whether there is a direct response from an administrator, whether the information matches, whether there is cross-verification, and the recency of the information, it can reduce the problem of incorrect information or temporary detection results being directly reflected in the operation data. Accordingly, the service robot and the collective intelligence-based multi-robot collaboration method of the present disclosure, which utilizes a RAG-based service policy updated with information collected by the service robot, can improve service quality tailored to actual store conditions.

[0169] Furthermore, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure converts store operation data into a RAG-based service policy reconstructed into searchable knowledge information, application condition information, and service strategy information, and accumulates the said RAG-based service policy on a main server; thus, field information acquired by individual service robots can be converted into a common policy asset that can be jointly utilized by all service robots. The service robot and collective intelligence-based multi-robot collaboration method of the present disclosure can store assigned area identification information, collection time information, service robot identification information, policy version information, update reason information, and reliability information together on the main server, thereby enabling traceable management of the policy's source, recency, and scope of application. Accordingly, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure can perform consistent guidance and decision-making anytime and anywhere within the store.

[0170] Furthermore, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure allows each service robot to access the RAG-based service policy of the main server and dynamically determine traversal strategies, response strategies, interaction strategies, work strategies, and collaboration strategies, thereby enabling the performance of services with higher situational adaptability than fixed rule-based guidance. The service robot and collective intelligence-based multi-robot collaboration method of the present disclosure can adjust movement paths, response content, response priority, the need for reconfirmation, and the support status of adjacent robots by reflecting the latest policies such as event information, congestion information, out-of-stock information, detour information, and administrator request information, thus enabling real-time response to changes in the store environment and customer demands. Accordingly, the service robot and collective intelligence-based multi-robot collaboration method of the present disclosure enables all multi-robots to perform cooperative and efficient services based on the same latest service policy.

[0171] However, as this has been explained above, a redundant explanation thereof will be omitted.

[0172] Using the embodiments of the present invention described above, those skilled in the art will be able to easily make various changes and modifications within the scope of the essential characteristics of the present invention. The content of each claim of the patent claims may be combined with other claims that are not related by reference within the scope of what can be understood from this specification.

Claims

Claim 1 A multi-robot collaboration method for updating a RAG-based service policy using collected information from each service robot comprises: a step in which at least one service robot circulates a pre-set assigned area within a store; a step in which the at least one service robot searches the assigned area to collect store information; a step in which the at least one service robot analyzes the store information to update store operation data corresponding to the store information; a step in which the at least one service robot converts the store operation data into a RAG-based service policy; and a step in which the at least one service robot uploads the RAG-based service policy to a main server. A multi-robot collaboration method comprising: a step in which at least one service robot accesses the RAG-based service policy and determines a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy; and, in the step of collecting store information, the at least one service robot actively queries a store manager within the assigned area and, based on the response to the query, collects interactive store information including at least one of event information, discount information, out-of-stock information, restocking schedule information, congestion information, detour information, and customer guidance priority information. Claim 2 A multi-robot collaboration method according to claim 1, wherein, in the step of circulating the pre-set assigned area, the at least one service robot generates a movement path based on path information of the assigned area including at least a part of the entire store area, and repeatedly circulates the assigned area based on a circulating schedule, movement priority, stopping point, re-search cycle, and avoidance path corresponding to the assigned area. Claim 3 delete Claim 4 A multi-robot collaboration method according to claim 1, wherein in the step of collecting store information, the at least one service robot collects non-interactive store information by detecting the shelf status, product display density, event banner exposure, customer dwell density, queue formation, passage availability, and staff placement status of the assigned area. Claim 5 A multi-robot collaboration method according to claim 1, wherein, in the step of collecting store information, at least one service robot generates composite store information for the same assigned area by combining conversational store information and non-conversational store information, and calculates a reliability score C of the composite store information using the following [Formula 1]. [Formula 1] C = w1*M + w2*V + w3*X + w4*R (wherein C is the reliability score, M is a value indicating whether the store manager has directly verified it, V is a value indicating whether the conversational collection result and the non-conversational collection result match, X is a value indicating whether there is cross-verification among multiple service robots, R is a value indicating the recency score based on the collection time, and w1, w2, w3, and w4 are weights.) Claim 6 A multi-robot collaboration method according to claim 1, wherein in the step of updating the store operation data, the at least one service robot analyzes the store information by performing at least one of text conversion of voice recognition data, classification of object recognition data, mapping with location information, determination of event occurrence, linkage of time information, identification of assigned area, and determination of priority. Claim 7 A multi-robot collaboration method according to claim 1, wherein in the step of updating the store operation data, the at least one service robot extracts operation information including at least one of product location information, event schedule information, discount application information, out-of-stock information, restocking schedule information, congested area information, detour requirement information, customer precautions information, manager delivery information, and guidance priority information from the store information, and updates the store operation data by comparing the operation information with previously stored store operation data when a preset update condition is satisfied. Claim 8 A multi-robot collaboration method according to claim 1, wherein the RAG-based service policy includes knowledge information searchable when determining a customer query or a service to be provided, application condition information serving as a criterion for determining whether to provide a service or the method of providing a service, and the service strategy defining the execution direction for each service strategy, wherein in the step of converting the store operation data into the RAG-based service policy, the at least one service robot classifies the store operation data into entity information, assigned area information, time information, query type information, and strategy type information, and generates the knowledge information, the application condition information, and the service strategy by recombining the classified store operation data based on information directly used for customer response, information determining whether to provide a service, and information used for performing a service. Claim 9 A multi-robot collaboration method according to claim 1, wherein in the step of uploading the RAG-based service policy to a main server, the at least one service robot uploads the RAG-based service policy to the main server based on at least one of a customer urgent request, identification of out-of-stock status, exceeding of a standard congestion level, a preset cycle, and a store manager command, and the main server stores responsible area identification information, collection time information, service robot identification information, policy version information, update reason information, and reliability information together with the RAG-based service policy. Claim 10 A multi-robot collaboration method according to claim 1, wherein at least one service robot accesses the RAG-based service policy to search for a policy reflecting the status of the assigned area, congestion, event status, and the operation status of other service robots, and determines the circulation strategy by determining at least one of the re-inspection frequency, bypass area, movement path, and circulation priority area based on the priority score and reliability score of the searched policy. Claim 11 A multi-robot collaboration method according to claim 1, wherein at least one service robot searches for a RAG-based service policy including a query type tag corresponding to a customer query, and determines a response strategy by determining a response content based on at least one of a response template identifier, priority score, product location information, discount application information, out-of-stock information, congestion area information, and detour need information included in the searched policy. Claim 12 A multi-robot collaboration method according to claim 1, wherein at least one service robot searches for a RAG-based service policy corresponding to whether a customer is approaching, whether an administrator is requesting, time zone, congestion level, and customer type, and determines a response strategy by determining at least one of whether to provide preemptive guidance, priority of response to a customer or administrator, level of explanation, and tone of voice of response based on the searched policy. Claim 13 A multi-robot collaboration method according to claim 1, wherein at least one service robot accesses the RAG-based service policy to search for a policy reflecting at least one of information discrepancy, lack of reliability, lack of stock, change in event status, and need for manager verification, and determines the need for at least one of re-search, re-verification, store manager query, event banner verification, and repetitive collection based on the searched policy, thereby determining the work strategy. Claim 14 A multi-robot collaboration method according to claim 1, wherein at least one service robot accesses the RAG-based service policy to search for a policy that reflects the status of the assigned area, congestion status, response status, support request information, and customer handover requirement information of another service robot, and determines the cooperation strategy by determining at least one of role sharing, information re-verification request, adjacent area support request, customer handover, and temporary joint patrol based on the searched policy. Claim 15 A service robot that implements real-time multi-robot collaboration by updating a RAG-based service policy with collected information, comprising: a memory for storing at least one command; and at least one processor for executing said at least one command, wherein the at least one processor is configured to circulate a pre-set assigned area within a store, is configured to collect store information by exploring said assigned area, is configured to update store operation data corresponding to said store information by analyzing said store information, is configured to convert said store operation data into a RAG-based service policy, is configured to upload said RAG-based service policy to a main server, and is configured to determine a service strategy including at least one of a circulating strategy, a response strategy, a response strategy, a work strategy, and a cooperation strategy by accessing said RAG-based service policy, wherein the at least one processor actively queries a store manager within said assigned area and collects interactive store information including at least one of event information, discount information, out-of-stock information, restocking schedule information, congestion information, detour information, and customer guidance priority information based on the response to said query.

Images