LLM-based intelligent warehouse management system and operating method thereof

Abstract

An LLM-based intelligent warehouse management system according to an embodiment of the present invention comprises: a cloud server; an edge device for monitoring a work performance situation in real time; and an on-premise server that responds to an exceptional situation by using stored information when the exceptional situation occurs, wherein the cloud server comprises: an LLM service unit that receives an input of a work instruction in natural language from a manager client and interprets the input work instruction by using LLM; an integrated database unit for storing work data to consider a warehouse situation; an intelligent service unit for allocating work according to a work plan, wherein the LLM-based intelligent warehouse management system comprises an AMR for performing work while autonomously moving according to the allocated work.

Description

LLM-based intelligent warehouse management system and its operation method

[0001] The present invention relates to a warehouse management system, and more specifically, to an LLM-based intelligent warehouse management system and a method of operation thereof that enables intuitive work instructions to transfer robots using a large-scale language model and improves the automation performance of artificial intelligence through continuous data collection to enhance the efficiency and safety of warehouse operations.

[0002] The present invention has been filed as the result of the following research project.

[0003] Research Project Name: 2025 DNA Large-Medium-Small Partnership Joint Entry

[0004] Research Project Title: Overseas Pilot Implementation and Market Expansion of DX Solutions and Logistics Platform Services Utilizing AI Technology

[0005] Project ID: G0702-24-1023

[0006] Research Management Agency: National Information & Communication Technology Promotion Agency (NIPA)

[0007] Warehouse Management Systems (WMS) are being designed to enhance the efficiency of logistics and warehouse operations in the retail industry, which is growing due to the digital transformation of logistics and the expansion of platform businesses. Generally, WMSs provide functions such as inventory management, order processing, and inbound / outbound management, and are systems that particularly focus on automating processes and minimizing manual labor.

[0008] Since most warehouse management systems are built based on fixed user interface specifications from existing systems, they require control using complex command systems. This approach degrades the user experience and ultimately ensures that only skilled personnel can use the system effectively.

[0009] In this regard, there is a recent warehouse management system designed to collect data generated in the warehouse and perform machine learning (Korean Patent Publication No. 10-2020-0048791 of the prior art), but there are limitations in learning in real time and making immediate decisions. In addition, since it requires a data processing process, it reduces the response speed when unexpected situations occur, thereby lowering the efficiency of warehouse operations.

[0010] Meanwhile, the warehouse environment is also a place where safety accidents are highly likely to occur due to the interaction between people and means of transport and environmental factors. Although a warehouse management system including safety management functions (Korean Patent Publication No. 10-2024-0003604 of the prior art) has been presented, it is limited to passive response and thus has limitations in detecting risk factors and executing measures in real time. In other words, conventional warehouse management systems lack mechanisms to continuously improve system performance based on user feedback or new data in terms of continuous system improvement and learning functions, which can lead to a decline in system efficiency over time.

[0011] To address these issues, there is a need for a warehouse management system that enables automatic decision-making through AI-based intuitive user instructions and real-time data analysis, and whose underlying integrated knowledge base and continuous learning mechanisms can be enhanced through the combination of cloud and edge computing technologies.

[0012] To solve the aforementioned problems, the technical objective of the present invention is to provide a warehouse management interface that performs tasks intuitively based on natural language-based instructions.

[0013] Furthermore, the technical objective of the present invention is to construct a warehouse management system capable of scalability and real-time processing by combining cloud computing and edge computing, wherein system performance is continuously improved through a knowledge base-based continuous learning mechanism.

[0014] Furthermore, the technical objective of the present invention is to provide a warehouse management system that achieves operational efficiency and promotes safety in the work environment through real-time data analysis and AI-based predictive decision-making.

[0015] An intelligent warehouse management system based on Large Language Models (LM) according to an embodiment of the present invention for achieving the above technical objectives comprises: a cloud server; an on-premises server that communicates with the cloud server and is deployed in at least one of the managed warehouses to receive and store real-time data from the site; an edge device that receives work guides from the cloud server through the on-premises server and monitors the status of work execution in real time; and an Autonomous Mobile Robot (AMR) that moves autonomously and performs work according to the work guides. The cloud server comprises: an LLM service unit that receives work instructions in natural language from an administrator client and interprets the work instructions using LLM; an integrated database unit that stores knowledge data regarding work and safety and information transmitted by the edge device; and an intelligent service unit that performs data analysis and decision-making to coordinate and optimize the operation of the entire warehouse management system based on artificial intelligence.

[0016] In addition, a method for operating an LLM (Large Language Models)-based intelligent warehouse management system according to an embodiment of the present invention for achieving the above technical objectives comprises: receiving work instructions in natural language from a manager client; interpreting the input work instructions using LLM; assigning tasks to each AMR (Autonomous Mobile Robot) based on the interpreted work instructions and warehouse conditions and transmitting the work instructions; having the AMRs move autonomously and perform the tasks according to the assigned tasks; and monitoring the work performance status of the AMRs in real time to verify the results of the work performance.

[0017] According to an embodiment of the present invention, an LLM-based intelligent warehouse management system can provide a warehouse management interface that performs tasks intuitively based on natural language instructions.

[0018] According to an embodiment of the present invention, the LLM-based intelligent warehouse management system has the effect of continuously improving performance through a continuous learning mechanism based on an integrated knowledge base, and enabling scalability and real-time response through the combination of cloud computing and edge computing.

[0019] According to an embodiment of the present invention, an LLM-based intelligent warehouse management system can increase the efficiency of warehouse operations and the safety of the work environment through real-time data analysis and decision-making predicted by AI.

[0020] The effects of the present invention are not limited to the effects described above, and should be understood to include all effects that can be inferred from the composition of the invention described in the description or claims of the present invention.

[0021] FIG. 1 is a drawing illustrating an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0022] FIG. 2 is a block diagram illustrating the configuration of a cloud server (200) according to one embodiment of the present invention.

[0023] FIG. 3 is a block diagram illustrating the detailed configuration of an intelligent service unit (230) according to one embodiment of the present invention.

[0024] FIG. 4 is a block diagram illustrating the configuration of an on-premises server (300) according to one embodiment of the present invention.

[0025] FIG. 5 is a block diagram illustrating the configuration of an edge device (400) according to one embodiment of the present invention.

[0026] FIG. 6 is a flowchart of a natural language-based work instruction method of an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0027] FIG. 7 is a flowchart of a work monitoring reporting method of an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0028] FIG. 8 is a flowchart of a safety management method of an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0029] The present invention will be described below with reference to the attached drawings. However, the present invention may be implemented in various different forms and is therefore not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification have been given similar reference numerals.

[0030] Throughout the specification, when it is stated that a part is "connected (connected, in contact, combined)" with another part, this includes not only cases where they are "directly connected," but also cases where they are "indirectly connected" with other members interposed between them. Furthermore, when it is stated that a part "includes" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but rather allows for the inclusion of additional components.

[0031] The terms used in this specification are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise.

[0032] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should not be understood as precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0033] Embodiments of the present invention will be described in detail below with reference to the attached drawings.

[0034]

[0035] FIG. 1 is a drawing illustrating an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0036] Referring to FIG. 1, the intelligent warehouse management system (100) may include an administrator client (10), a cloud server (200), an on-premises server (300), an edge device (400), an IoT sensor (30), and an AMR (40).

[0037] First, the administrator client (10) may include an administrator and management-side warehouse management software and applications. The administrator client (10) may transmit and receive information within a system including a cloud server (200), an on-premises server (300), and an edge device (400).

[0038] The cloud server (200) can acquire and store large-scale data and perform big data analysis and model prediction. It can collect and analyze all data from the warehouse to optimize work patterns or generate prediction models, and can centrally manage the system even from a distance. This is due to the scalability of the cloud, which makes it possible to connect multiple various sensors and devices and integrate data.

[0039] The cloud server (200) can obtain natural language instructions and knowledge data regarding the work from the administrator client (10) and store them as integrated data.

[0040] Here, natural language instructions include all instructions generated in natural language for the performance of a task, and knowledge data may be the entirety of work-related data and documents, including manuals, safety guidelines, and historical work data.

[0041] The cloud server (200) can generate work instructions by processing natural language instructions with a large-scale language model, receive data from edge computing devices, and optimize predictions for warehouse management.

[0042] The on-premises server (300) is a server deployed in at least one of the warehouses managed by the intelligent warehouse management system and can support real-time operations such as immediate data processing and safety management.

[0043] The on-premises server (300) includes local database and cache server functions, and can transmit or synchronize data collected on-site to the server.

[0044] Additionally, the on-premises server (300) enables the core functions of the intelligent warehouse management system (100) to be performed even when communication with the cloud server (200) is limited or delayed.

[0045] For example, the on-premises server (300) can continue real-time operations in conjunction with the edge device (400) even in situations of network delay or inability to access the cloud.

[0046] The edge device (400) can collect and manage warehouse data through various sensors, including monitoring devices and environmental sensors, and transmit or synchronize the collected data to the cloud server (200). The cloud server (200) receives the collected data, performs AI-based decision-making, and as a result, can transmit work instructions to the edge device (400). These work instructions can be delivered to workers and AMRs (Autonomous Mobile Robots) and utilized for warehouse operations.

[0047] In addition, the edge device (400) can continue real-time operations in conjunction with the on-premises server (300) even in situations where network delays or cloud access are unavailable.

[0048] The edge device (400) can detect safety conditions at the work site in real time through CCTV cameras and various sensors, and can generate notifications or provide response measures when a dangerous situation occurs.

[0049] For example, the edge device (400) can monitor warehouse environment data such as temperature, humidity, and air pollution levels, and send a notification to the manager when the threshold value is exceeded.

[0050] In addition, the edge device (400) can perform the role of efficiently managing and controlling the AMR (40) in the warehouse.

[0051] The IoT sensor (30) is a CCTV camera and various sensors installed in a warehouse, and may be a sensor system including multiple cameras and sensing equipment.

[0052] AMR (40) is a mobile robot equipped with a communication function that transmits and receives signals and information with sensors and edge devices, and may be a fleet composed of multiple robots.

[0053] FIG. 2 is a block diagram illustrating the configuration of a cloud server (200) according to one embodiment of the present invention.

[0054] The cloud server (200) can acquire and store large-scale data and perform big data analysis and model prediction. It can collect and analyze all data from the warehouse to optimize work patterns or generate prediction models, and can centrally manage the system even from a distance. This is due to the scalability of the cloud, which makes it possible to connect multiple various sensors and devices and integrate data.

[0055] Referring to FIG. 2, the cloud server (200) may include an integrated database (210), an LLM service unit (220), an intelligent service unit (230), and an API gateway (240).

[0056] The integrated database (210) can store knowledge data and obtain and store information from the edge device (400).

[0057] For example, knowledge data including work manuals, safety guidelines, past work data, etc. can be stored, and data from edge devices (400), such as various sensor data and work data, can be stored.

[0058] In addition, the integrated database (210) is equipped with version control and real-time update functions and can search for and store context-based information in conjunction with LLM.

[0059] Large Language Models (LMs) are deep learning language models that are pre-trained with a vast amount of data, and the LLM service unit (220) can host and run existing large language models.

[0060] The LLM service unit (220) can interpret natural language instructions from the administrator client (10) and perform language processing to generate natural language instructions.

[0061] In addition, the LLM service department (220) can receive knowledge data from the integrated database (210), perform context analysis and language processing, and generate more accurate natural language instructions by reflecting this.

[0062] The intelligent service unit (230) can perform data analysis and decision-making to coordinate and optimize the operation of the entire system based on artificial intelligence. Here, automatic decision-making regarding predictive analysis, optimal path setting, task scheduling, and anomaly detection is performed.

[0063] In addition, the intelligent service unit (230) is designed in a modular manner, so that it can expand its functions by adding new distributed learning and inference models.

[0064] The intelligent service unit (230) will be described in detail below with reference to FIG. 3.

[0065] The API gateway (240) can efficiently manage the flow of data between various service modules by mediating communication with external systems.

[0066] For example, the API gateway (240) can be integrated with external logistics systems such as ERP (Enterprise Resource Planning) and TMS (Transportation Management System) to support the linkage of the entire supply chain.

[0067] In addition, the API gateway (240) maintains security through authentication and authorization management and reduces system load by optimizing data requests.

[0068]

[0069] FIG. 3 is a block diagram illustrating the detailed configuration of an intelligent service unit (230) according to one embodiment of the present invention.

[0070] Referring to FIG. 3, the intelligent service unit (230) may include a central control engine (231), a task scheduler module (232), an AI prediction module (233), and a digital twin module (234).

[0071] First, the central control engine (231) can continuously monitor and improve the performance of the system based on warehouse operation data.

[0072] The central control engine (231) plays a key role in the system for coordinating and optimizing various operations of the warehouse, and can establish performance evaluation indicators and make decisions such as work plans and safety measures.

[0073] The task scheduler module (232) performs task scheduling based on the established task plan.

[0074] Here, the task scheduler module (232) can distribute work resources and assign tasks to be performed by the AMR (40) and the worker (20).

[0075] The AI ​​prediction module (233) can predict risk factors in work situations by utilizing various machine learning models.

[0076] The AI ​​prediction module (233) can identify work patterns and performance through big data analysis of all data stored in the integrated database (210).

[0077] Here, the AI ​​prediction module (233) can obtain performance data and feedback from the manager client (10) to derive improvements and perform periodic updates of the machine learning model.

[0078] The digital twin module (234) can collect data in real time and reflect it on a virtual screen to create an environment identical to the actual work site.

[0079] Here, work scenarios can be generated by reflecting all real-world data in real time, and virtual workspaces can be implemented through various types of simulations.

[0080] Additionally, the digital twin module (234) may include a feedback loop that performs periodic updates by applying manager feedback after the implemented simulation is provided to the manager.

[0081] To this end, the intelligent warehouse management system (100) needs to collect all data in real time and continuously update the learning model, and a virtual simulation similar to the actual system can be provided.

[0082] FIG. 4 is a block diagram illustrating the configuration of an on-premises server (300) according to one embodiment of the present invention.

[0083] Referring to FIG. 4, the on-premises server (300) may be configured to include a local database (310), a simulation provider (320), a work guide provider (330), and a monitoring display (340).

[0084] The on-premises server (300) is a physical server infrastructure deployed on-site that can receive and store real-time data.

[0085] The on-premises server (300) acts as an intermediate point between the cloud server (200) and the edge device (400) and can manage data that requires immediate access and frequently used data caching.

[0086] The local database (310) can collect and store all data generated at the work site.

[0087] The local database (310) can collect data generated in real time and historical data, and obtain and store data on whether the AMR (40) and the worker (20) have performed work and work performance data.

[0088] In addition, the local database (310) can supplement the stability of the system by responding to real-time data transmission, etc., even in the event of an unexpected communication accident.

[0089] The simulation providing unit (320) can provide a virtual simulation to the administrator client (10). The simulation providing unit (320) provides a virtual work simulation using optimal work information derived by the cloud server (200) continuously learning all data occurring at the actual work site.

[0090] The work guide providing unit (330) can provide the work plan and scheduling generated by the cloud server (200) to the edge device (400). Through this, the edge device (400) can deliver accurate work plans and scheduling to the worker (20) and AMR (40) and control the work of the worker (20) and AMR (40).

[0091] The monitoring display unit (340) can display monitoring data from the device (400) to the administrator client (10) on the screen.

[0092] The monitoring display unit (340) may include graphs, charts, dashboards, etc. that visually provide real-time and historical monitoring data, and through this, the administrator client (10) can identify the current status and view past operations.

[0093] FIG. 5 is a block diagram illustrating the configuration of an edge device (400) according to one embodiment of the present invention.

[0094] The edge device (400) is a device that performs data processing at the site and near the site, and can be implemented in various forms.

[0095] For example, the edge device (400) may be a terminal including an IoT sensor, a smart camera, uCPE equipment and a server, a processor and communication functions.

[0096] Referring to FIG. 5, the edge device (400) may include a sensor control unit (410) and an AMR control unit (420).

[0097] The sensor control unit (410) can obtain information in real time through the communication function with the sensors mounted on the IoT sensors (30) and AMR (40) installed throughout the warehouse (50).

[0098] The sensor control unit (410) can not only obtain sensor information but also control the sensor immediately.

[0099] The sensor control unit (410) may be composed of an environment sensor module (411), an AMR position sensing module (412), and a work monitoring module (413).

[0100] The environment sensor module (411) can detect environmental information such as the temperature, humidity, and air pollution level of the warehouse (50) in real time.

[0101] Here, the environment sensor module (411) can be managed through sensor control to maintain optimal storage conditions in the warehouse based on the detected results.

[0102] For example, the environment sensor module (411) can automatically activate the heating system and ventilation system when the temperature or humidity of the work site reaches a dangerous level based on sensor data.

[0103] The AMR location sensing module (412) can detect the location of the AMR (40) and detect whether it has deviated from the path within the work area.

[0104] Additionally, the AMR location sensing module (412) can perform emergency safety measures based on the detected result.

[0105] For example, the AMR location sensing module (412) may control the AMR (40) to stop immediately or select an alternative route, and may suspend all operations in the work area until the danger is resolved.

[0106] The work monitoring module (413) can monitor the overall work status in real time. Here, the work monitoring module (413) can receive sensed data and detect and respond to dangerous situations occurring in the work environment.

[0107] In addition, the work monitoring module (413) can distinguish dangerous situations by performing analysis using machine vision and AI algorithms, and can execute immediate warnings and notifications for urgent dangerous situations and automatically perform safety measures through control regarding dangerous situations.

[0108] Additionally, the work monitoring module (413) can evaluate the performance of the AMR (40) and the worker (20) through monitoring.

[0109] For example, a performance evaluation can be derived using key performance indicators (KPIs) and actual performance relative to targets as indicators, and such monitoring and performance evaluation data can be provided to a cloud server (200) for retraining.

[0110] In addition, all results of the executed work can be automatically transmitted to the administrator client (10), cloud server (200), and on-premises server (300).

[0111] The AMR control unit (420) may be composed of a task assignment module (421), a path planning module (422), and a task prediction module (423).

[0112] The work assignment module (421) can assign appropriate work to the AMR (40) and the worker (20) according to the workload.

[0113] Additionally, the work assignment module (421) may include a function to dynamically adjust work based on the availability of work resources, including AMR (40) and worker (20).

[0114] The path planning module (422) can set a path so that the AMR (40) finds the optimal path and moves within the warehouse. Here, the path planning module (422) can set a path by receiving the optimal path derived through a path search algorithm from the cloud server (200) and drive the AMR (40) along the optimal path accordingly.

[0115] The work prediction module (422) can predict the work estimated time and path efficiency of the AMR (40).

[0116] Here, the work prediction module (422) can optimize future work plans through AI-based analysis, thereby preventing work from becoming stagnant or causing work bottlenecks in advance.

[0117] FIG. 6 is a flowchart of a natural language-based work instruction method of an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0118] Referring to FIG. 6, in step (S110), the cloud server (200) receives natural language instructions from the administrator client (10).

[0119] In this case, natural language instructions include text and voice data composed of natural language and may include corresponding context information.

[0120] In step (S120), the cloud server (200) can interpret the natural language instruction and process it through a large-scale language model of the LLM service to generate an appropriate work instruction.

[0121] In step (S130), the cloud server (200) can learn knowledge data including work manuals, safety guidelines, past work data, etc., and analyze the learned results to establish an optimal work plan.

[0122] In step (S140), the cloud server (200) performs work scheduling based on the established work plan. At this time, the cloud server (200) can predict work performance by resource using artificial intelligence and assign tasks to be performed by the AMR (40) and the worker (20).

[0123] In step (S150), the cloud server (200) transmits work instructions to the AMR (40) and the worker (20) to perform the assigned work.

[0124] At this time, the work instruction is generated in natural language from the cloud server (200) and transmitted to the edge device (400).

[0125] Next, in step (S160), the edge device (400) can control the AMR (40) to perform a task according to task scheduling and task instructions.

[0126] FIG. 7 is a flowchart of a method for performing and reporting work monitoring of an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0127] Referring to FIG. 7, in step (S161), the edge device (400) can track the work status in real time to monitor the work progress of the AMR (40) and the worker (20).

[0128] In step (S162), the cloud server (200) can analyze the data from the monitoring step to verify the completion of the task and the results of the task execution.

[0129] For example, it is possible to verify whether the AMR (40) has completed the assigned tasks and instructions according to the established work plan and the results of the execution.

[0130] In step (S163), the cloud server (200) can analyze and evaluate the performance of the worker (20) and the AMR (40) through the verification above.

[0131] At this time, the cloud server (200) can evaluate performance based on key performance indicators (KPIs) and actual performance compared to targets, and can also perform evaluation by obtaining relevant data from the edge device (400).

[0132] Additionally, the cloud server (200) can classify the AMR (40) into multiple AMRs and AMR fleets into team units, analyze performance metrics, and evaluate performance capabilities by team unit.

[0133] The above team-based evaluation can be performed by classifying work teams according to work time, work radius, and work tasks to form one or more work teams, and by analyzing performance metrics for each work team.

[0134] In step (S164), the cloud server (200) can store performance evaluation data in a database and retrain to derive system optimization.

[0135] In step (S165), the cloud server (200) reports the optimization prediction results to the administrator client (10) and can receive feedback from the administrator client (10).

[0136] In step (S166), the cloud server (200) can update the system by storing the feedback in a database and retraining it. Here, work efficiency can be predicted through big data analysis and machine learning, and a prediction scenario can be generated and simulated.

[0137] FIG. 8 is a flowchart of a safety management method of an intelligent warehouse management system (100) according to one embodiment of the present invention.

[0138] Referring to FIG. 8, in step (S171), the edge device (400) can monitor the working environment of the intelligent warehouse management system (100) in real time and perform analysis of monitoring data and sensor data.

[0139] In step (S172), the edge device (400) can detect dangerous situations occurring in the work environment through monitoring data and sensor data. Additionally, it can distinguish dangerous situations by performing analysis using machine vision and AI algorithms.

[0140] In step (S173), the edge device (400) can perform an immediate response to the detected dangerous situation. The detected dangerous situation is transmitted to the administrator client (10), and the edge device (400) can generate immediate warnings and notifications depending on the dangerous situation.

[0141] Warnings and notifications of such edge devices (400) are generated by auditory, visual, vibration means and one or more other means that are generated immediately, and can be delivered to a worker (10) for emergency response.

[0142] In step (S174), the edge device (400) can automatically perform safety measures through control regarding dangerous situations.

[0143] For example, the edge device (400) may control the AMR (40) to stop immediately or select a detour route, and may suspend all operations in the work area until the risk is resolved.

[0144] Additionally, the edge device (400) can ensure safety by automatically adjusting the speed of the AMR (40) or modifying the path when the AMR (40) and the worker (20) get closer to each other than a set amount.

[0145] Additionally, the edge device (400) can automatically activate the heating system and ventilation system based on sensor data when the temperature or humidity of the work site reaches a dangerous level.

[0146] In step (S175), the edge device (400) may provide the result executed according to the automatic safety measure. Then, after the dangerous situation is resolved, whether the automatic safety measure was executed and the result executed are returned to the edge device (400).

[0147] In addition, the execution status and the result of the execution can be automatically transmitted to the administrator client (10), the cloud server (200), and the on-premises server (300).

[0148] At this time, the automatically transmitted information may include detected risks, actions taken, and the results of the actions, and the LLM service unit (220) of the cloud server (200) may generate a natural language report and provide it to the administrator client (10).

[0149] By relearning the execution results in this way, the intelligent warehouse management system (100) can continuously analyze the causes of problems and establish future predictions and response measures.

[0150] The method according to the embodiments of the present invention described above may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the computer-readable recording medium may be specially designed and configured for the embodiments of the present invention, or may be known and available to a person skilled in the art of computer software. The computer-readable recording medium includes hardware configured to store and execute program instructions, such as magnetic recording media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; ROMs; RAMs; and flash memory. Program instructions include machine code generated by a compiler and high-level language code that can be executed on a computer using an interpreter. The hardware may be configured to operate as one or more software modules to process the method according to the present invention, and vice versa.

[0151] The method according to an embodiment of the present invention may be executed in the form of program instructions on an electronic device. The electronic device includes portable communication devices such as smartphones or smartpads, computer devices, portable multimedia devices, portable medical devices, cameras, wearable devices, and home appliances.

[0152] The method according to an embodiment of the present invention may be provided as included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable recording medium or online through an application store. In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily created in a storage medium such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0153] Each component, such as a module or a program, according to an embodiment of the present invention may be composed of a single or multiple sub-components, and some of these sub-components may be omitted or additional sub-components may be included. Some components (modules or programs) may be integrated into a single entity and may perform the functions performed by each corresponding component prior to integration in the same or similar manner. Operations performed by a module, program, or other component according to an embodiment of the present invention may be executed sequentially, in parallel, iteratively, or heuristically, or at least some operations may be executed in a different order, omitted, or additional operations may be added.

[0154] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0155] The scope of the present invention is defined by the claims set forth below, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention.

[0156] The modes for carrying out the invention are described together in the best mode for carrying out the invention.

[0157] An LLM-based intelligent warehouse management system according to an embodiment of the present invention can provide a warehouse management interface that performs tasks intuitively based on natural language instructions.

[0158] The LLM-based intelligent warehouse management system according to an embodiment of the present invention can increase the efficiency of warehouse operations and the safety of the work environment through real-time data analysis and decision-making predicted by AI.

Claims

1. In an LLM (Large Language Models)-based intelligent warehouse management system, Cloud server; An on-premise server that communicates with the above-mentioned cloud server and is deployed in at least one of the managed warehouses to acquire and store real-time data from the site; An edge device that receives a work guide from the cloud server through the on-premises server and monitors the status of work execution in real time; and An AMR (Autonomous mobile robot) that moves autonomously and performs tasks according to the above work guide; Includes, The above cloud server comprises an LLM service unit that receives work instructions in natural language from an administrator client and interprets the work instructions using LLM; An integrated database unit that stores knowledge data regarding work and safety and information transmitted by the edge device; and An intelligent service unit that performs data analysis and decision-making to coordinate and optimize the operation of the entire warehouse management system based on artificial intelligence; including LLM-based intelligent warehouse management system.

2. In Paragraph 1, The edge device monitors the AMR, provides a work result after the work is completed, and evaluates performance by analyzing the work result. LLM-based intelligent warehouse management system.

3. In Paragraph 1, The above LLM service unit generates the above AMR's work report in natural language, and Providing the above-generated work report to the above-mentioned administrator client LLM-based intelligent warehouse management system.

4. In Paragraph 1, The above intelligent service department A central control engine that makes decisions such as work plans and safety measures based on warehouse operation data; A task scheduler module that assigns tasks according to the above task plan; AI prediction module for predicting risk factors in work situations; and A digital twin module that collects data in real time and reflects it on a virtual screen to replicate the actual work site; including LLM-based intelligent warehouse management system.

5. In Paragraph 4, The above digital twin module is By reflecting real-time collected data in a virtual environment, it is identical to the actual work site. It is to provide the implemented simulation to the administrator client and perform periodic updates by applying administrator feedback. LLM-based intelligent warehouse management system.

6. In Paragraph 1, The edge device described above includes an environment sensor module, an AMR location sensing module, and a task monitoring module. The above-mentioned task monitoring module is, It analyzes sensor data to detect operational risks, executes emergency alerts, controls, and safety measures according to the hazardous situation, and provides the execution results to the aforementioned administrator client. LLM-based intelligent warehouse management system.

7. In Paragraph 6, The above task monitoring module performs periodic updates by applying feedback from the administrator client. LLM-based intelligent warehouse management system.

8. In the operation method of an LLM (Large Language Models)-based intelligent warehouse management system, Step of receiving work instructions in natural language from the administrator client; A step of interpreting the above-mentioned work instructions using LLM; A step of assigning tasks to each AMR (Autonomous mobile robot) and transmitting work instructions based on the above-mentioned interpreted work instructions and warehouse conditions; A step in which the AMR moves autonomously and performs a task according to the above-mentioned assigned task; and A step of verifying the work execution result by monitoring the work execution status of the above AMR in real time; Operation method of an LLM-based intelligent warehouse management system.

9. In Paragraph 8, The step of verifying the results of the work execution by monitoring the above is, By analyzing the results of the above task execution after the task is completed, the performance of the above AMR It includes the evaluation step Operation method of an LLM-based intelligent warehouse management system.

10. In Paragraph 8, The step of verifying the results of the work execution by monitoring the above is, The method includes the step of generating a job report of the above AMR in natural language and providing the generated job report to the administrator client. Operation method of an LLM-based intelligent warehouse management system.

11. In Paragraph 8, The step of transmitting work instructions assigned to each AMR above is, A step of making decisions, such as work plans and safety measures, based on warehouse operation data; A step of assigning tasks according to the above work plan; AI prediction step for predicting risk factors in a work situation; and Collect data in real time and reflect it on a virtual screen A digital twin stage implemented identically to the actual work site; including Operation method of an LLM-based intelligent warehouse management system.

12. In Paragraph 11, The above digital twin step is, It provides administrator clients with a simulation implemented identically to the actual work site by reflecting real-time collected data in a virtual environment, and Performing periodic updates by applying feedback from the above administrator client Operation method of an LLM-based intelligent warehouse management system.

13. In Paragraph 8, The step of verifying the results of the work execution by monitoring the above is, Analyze sensor data to detect operational risks, and execute emergency alerts, controls, and safety measures based on hazardous situations, and the execution results The above-mentioned administrator client is Operation method of an LLM-based intelligent warehouse management system.

14. In Paragraph 13 The step of verifying the results of the work execution by monitoring the above is, Performing periodic updates by applying feedback from the above administrator client Operation method of an LLM-based intelligent warehouse management system.

Images