Autonomous Inventory Control System with Artificial Intelligence Orchestration

Abstract

The invention is an AI-powered image analysis and decision-making system that enables the automation of field inspection, anomaly detection, inventory counting, and verification processes. It utilizes images captured by field personnel via mobile devices to perform inventory detection, counting, verification, and anomaly analysis using multiple AI techniques, and integrates the results into corporate systems for reporting.

Description

1 TARIFF Autonomous Inventory Control System with Artificial Intelligence Orchestration Technical Area The invention involves 5 processes: field inspection, anomaly detection, inventory counting, and verification. with an AI-powered image analysis and decision-making system for automation It is related. State of the Art Today, organizations need to count, verify, and control their warehouse and field inventories. their operations mostly by manual methods or with limited automation tools This process typically involves field personnel conducting a headcount, counting, etc. by recording the results and documenting them with photographs when necessary These methods are being implemented. However, these methods are susceptible to human error and compromise the accuracy of counting and data. It includes many factors that reduce its reliability. 15 Today, the methods used in inventory counting and verification processes are basically threefold. They are divided into main categories. In manual counting and recording methods, field personnel physically count the inventories to be counted. They count them and manually record the results. The images are mostly It is used solely as documentation and is not directly integrated into the census process. 20 This approach is not implemented. The main shortcomings of this approach are; its susceptibility to human error, and its labor-intensive nature. The reason for this is that the census results are not verifiable. This situation arises... The main technical problem it created was that the images remained only as passive data and were not included in the inventory. The verification process cannot be performed automatically and reliably. Thanks to advancements in computer vision techniques in single-model-based object detection systems, 25 Inventory of objects from images is used by deep learning-based object detection models such as YOLO. Detection and counting can be done. However, these systems are generally only available when pre-trained. They are able to recognize the classes and fully understand the diversity in real field conditions. These systems are unable to meet the requirements. The main shortcomings of these systems are; those lacking in the training dataset. Inability to recognize objects, poor adaptation to different lighting, angles and positioning conditions, 30 The problem is the limited ability to conduct contextual analysis. This reduces the accuracy of the counting and a technical problem that limits the reliability of the system in real field environments It constitutes. In multimodule systems with a fixed workflow, multiple analysis modules can be used together. These modules are used, but they are executed in a fixed sequence. In this approach, every 35 The same analysis steps are applied to the image. The shortcomings of these systems are; image 2 The inability to dynamically manage the analysis sequence, which should change according to the content, is unnecessary. Increased processing time due to module execution, increased computational cost The rise is due to limited adaptation to different scenarios. This situation has led to... The fundamental technical problem is the inability to automatically determine image-specific analysis requirements, and This is due to the inefficient execution of the analysis process. 5 In conclusion, due to the negative aspects described above and the current solutions being the subject of discussion... Due to its shortcomings, it has become necessary to make improvements in the relevant technical field. Purpose of the Invention The invention represents a new breakthrough in this field, unlike the structures used in the current technology. It is a structure with different technical specifications that it brings together. The main purpose of the invention is field inspection, anomaly detection, inventory counting, and verification. AI-powered image analysis and decision-making that enables the automation of processes. The goal is to provide a delivery system. Another aim of the invention is to determine which analyses can be applied to the incoming images and these 15 an agentic artificial intelligence that can autonomously decide in what order the analyses should be run Intelligence is about providing decision-making mechanisms. Another aim of the invention is to create a powerful hybrid analysis by combining different artificial intelligence techniques. The aim is to provide an architecture that integrates inventory identification, counting, verification, and anomaly analysis. by performing this, it provides much higher accuracy compared to manual methods; 20 in hand storing the analysis results in a database and generating automated reports. to provide. To achieve the objectives described above, the invention, field inspection, anomaly detection, artificial intelligence that enables the automation of inventory counting and verification processes It is an image analysis and decision-making system supported by mobile devices for field personnel. 25 Using the images it captures, it applies multiple artificial intelligence techniques to the images. Performing inventory identification, counting, verification and anomaly analysis, and reporting the results obtained. It encompasses a software architecture that integrates with corporate systems and provides reporting. The structural and characteristic features and all the advantages of the invention are given in the figures below. Thanks to the detailed explanation written with references to the figures, 30 is clearer. This will be understood. Therefore, the evaluation should also take these figures and detailed explanations into account. It must be done by taking it. Figures that will help understand the invention. Figure 1 shows the general architecture of the system that is the subject of the invention. 35 3 Explanation of Part References 101. Field personnel 102. Mobile application 103. Agentic AI decision unit 104. Cloud server 5 105. Image analysis modules 106. Inventory and analysis database Application 107. Report 108 109. Manager 10 Detailed Description of the Invention This detailed explanation describes the AI ​​orchestration and autonomous inventory system that is the subject of the invention. The preferred structures of the audit system are solely for the purpose of better understanding the subject. It is explained as follows: 15 The invention relates to field inspection, anomaly detection, inventory counting, and verification processes. an AI-powered image analysis and decision-making system that enables automation using images taken by field personnel (101) via mobile devices, Inventory detection, counting, verification, and analysis of images using multiple artificial intelligence techniques. 20 It includes a software architecture that reports on this. The invention relates to warehousing and storage in telecommunications, energy, retail, manufacturing, logistics and similar sectors. It is applicable in all areas where field inventory management processes are involved. Furthermore... This system includes not only inventory counting but also field inspection, inventory verification, and material handling. 25 that can also be adapted to different application scenarios such as control and operational quality monitoring It has a flexible architecture. One of the most important innovations of the invention is which analyses are performed on the incoming images. It autonomously decides which analyses will be applied and in what order they will be run. It includes an agentic artificial intelligence decision-making mechanism capable of providing this. In this way; • The analysis process is dynamically adapted to the image content, 30 • Unnecessary calculation processes are reduced, • Processing time and system load are reduced, • Counting accuracy is improved. The invention creates a powerful hybrid analytics architecture by combining different artificial intelligence techniques. It offers. In this context; 35 • Training the YOLO model for objects that have no real-world equivalent. 4 • Unknown object identification and contextual analysis using Vision Language Model techniques, • Authentication via barcode / QR scanning, • Error detection using anomaly detection algorithms, • Inventory counting is carried out using the Vision Language Model. This hybrid structure increases the system's resilience to variations in real-world field conditions. It increases. The system integrates inventory identification, counting, verification, and anomaly analysis. This allows for much higher accuracy compared to manual methods. The analysis results obtained within the scope of the invention are stored in a database and are automated. Reports are being generated. This enables managers to make quick and accurate decisions. 10 It makes things easier. The invention involves imaging materials located in warehouse, field, or inventory environments. Agentic artificial intelligence decision-making enables automatic detection, counting, verification, and reporting. It is an end-to-end autonomous analysis system with a mechanism. The system; from image acquisition... The entire process, from report generation to completion, is automated without human intervention. 15 is carrying out. Field personnel (101) take a photograph of the inventory during the warehouse count. The photograph is used for the application. (102) is sent to the agentic artificial intelligence decision unit (103) via Agentic artificial intelligence. The decision unit (103) evaluates the content of the image and determines which analyses (105) will be applied. decides. These analyses (105); inventory determination, counting, anomaly detection, barcode / QR reading and 20 Vision-Language supported object detection. Analyses in line with Agentic decision (105) The process is executed, and the cycle continues until the resulting output provides sufficient accuracy. The census results are recorded in the database (106). This data includes the census history, the results of the survey, It includes anomaly records and performance data. Analysis results are transmitted to the application (107). Counting After completion, a report (108) is automatically generated and presented to the manager (109). 25 The working principle of the invention is as follows: The functions that will operate in sequence according to the invention are as follows; • Field personnel (101) check warehouse inventory time via mobile application (102) It takes a picture and sends the image to the system. • Images transmitted by the mobile application (102) are transferred to the cloud server (104) and analyzed The process is initiated. 30 • Agentic AI decision unit (103), which analyses (105) are performed on the incoming image will be implemented autonomously (inventory identification, counting, anomaly analysis, barcode / QR reading) He decides. • In accordance with the decision, image analysis (105) modules are run. The resulting agentic artificial The result is sent back to the intelligence decision unit (103). 35 until the result is satisfactory for the decision unit (103). The cycle continues. • The obtained analysis results are recorded in the database (106). • Analysis results are transferred to the application (107). • The counting report (108) is automatically generated by the system and the manager (109) is presented to the user.

Claims

6 REQUESTS 1. Field inspection, anomaly detection, inventory counting and verification processes AI-powered image analysis and decision-making that enables automation. It is a system, and its feature is; Images of inventories located in the field environment are obtained via mobile devices. 5 field personnel (101), mobile that transfers images obtained by field personnel (101) to the system application (102), by evaluating the content of the images received via the mobile application (102) Agentic artificial intelligence autonomously determines the types and order of analyses to be applied. intelligence decision unit (103), mobile application (102), agentic artificial intelligence decision unit (103), image analysis modules (105), inventory and analysis database (106), application (107) and report (108) cloud server providing data processing and communication infrastructure (104), inventory identification, object counting, verification, anomaly analysis, barcode / QR reading and 15 At least one of the object definition processes supported by the Vision Language Model. image analysis modules (105) the analysis results produced by the image analysis modules (105), counting history, An inventory and analytics database that stores anomaly records and performance data. (106), 20 application that transmits analysis results to the user or corporate system (107), inventory count, verification results, detected anomalies, and analysis performance report containing the outputs related to (108), Manager who viewed the report (108) and made a decision based on the inventory audit result. (109) 25 It includes.

Images