Configuration based on Industry 4.0 technologies as a step towards an affordable smart warehouse

1.1.Warehouse control and management systems

Inventory management is the process of ordering, handling, storing, and using a company’s non-capitalized assets, also known as its inventory. Inventory a complete list of items such as products, materials, or supplies that are stored in a warehouse before they are produced, shipped, or sold [23]. An inventory system is the set of policies and controls that monitors inventory levels and determines what levels should be maintained, when the inventory should be replenished, and what size the order should be, providing in turn the organizational and operational policies to maintain and track products in stock [11]. In an inventory system with continuous review, like Economic Order Quantity (EOQ) inventory methodology, the optimal inventory size is the one that minimizes its costs. EOQ, represents the ideal quantity of units a company should purchase to meet demand while minimizing inventory costs, in its formula, the specific point R, at which an order will be placed, is determined, as well as its size Q. The order point R is always a specific number of units. An order of size Q is placed when available inventory, currently in stock or on order, reaches point R [21].

Inventory control systems are technological solutions that integrate all aspects of an organization’s inventory tasks [20]. Modern warehouses use a warehouse management system (WMS) that can coexist with appropriate and functional associated systems for information exchange. It is important to highlight that the WMS is responsible for executing all the movements carried out in the warehouse, and must be compatible with other systems [16], which is why it is necessary for the implementation of technical automation solutions in warehouse operations, providing the technical framework for the connectivity of these technologies and supporting the day-to-day business in a warehouse [9], also facilitating its global management by allowing the accumulation of all operations derived from its processes, being the main source of concentration of information regarding its flows [5].

This paper addresses the challenges faced specially by small and medium-sized businesses (SMBs), which are the backbone of most economies, in managing their warehouses effectively through an automated solution that streamlines inventory control and demand forecasting, enabling real-time inventory accuracy and transforming the warehouse into a strategic asset that drives logistics efficiency and competitive advantage.

2.1.State of technique

The fiercely competitive landscape presents unique challenges for SMBs that arises to streamline warehouse performance given that demand is uncertain, and it can be difficult to predict demand accurately, especially for new products or products with historical data that is only available for a few months. Rapid growth, trends, and seasonality can also impact demand and lead times. This can translate into significant inventory management problems, such as excess or shortage of stock, high storage costs, and lost sales opportunities.

The efficacy of inventory management stands as a pivotal factor in meeting customer demands, curtailing costs, and preserving profitability [11]. Generally, the inventory system is carried out through a digital automation process, in most cases through a WMS, however, more specific systems are added to this to improve control over warehouse operations [16]. Where demand and budget allow, sophisticated software, robotics and automation are having a groundbreaking impact on logistics [21].

As previously stated, IoT refers to the interconnection of everyday objects through the Internet infrastructure [2]. In the context of inventory management, Industrial IoT (IIoT), can have a profound impact, facilitating better resource allocation and utilization. By seamlessly connecting RFID tracking devices, it empowers real-time inventory tracking and enhances product identification accuracy, ultimately leading to improved operational efficiency.

AI has demonstrated remarkable adaptability within the ever-evolving technological landscape. This is evidenced by the significant advancements in both technical complexity and the breadth of tasks AI can perform [19]. Moreover, its applicability has demonstrably expanded.

Within an automated warehouse environment, AI algorithms perform in-depth analyses on data acquired by IIoT sensors embedded in items. This analysis enables the identification of recurring patterns and trends within the data to generate forecasts of future product requirements. This capability facilitates proactive inventory management, ultimately influencing production planning optimizations.

Through the adept identification of patterns and trends hidden within vast data sets, AI demonstrably enhances the accuracy of demand forecasting, since supply chains often suffer from excessive inventory and stockout costs due to inaccurate demand forecasting [22]. Inventory position is defined as the quantity on hand plus the quantity ordered minus backlogged orders [4]. The solution for a fixed order quantity model may stipulate when the demand for the product is constant and uniform throughout the period, and the delivery time and the price per unit of the product is constant, factors that are considered for meticulously adjust inventory levels.

The rise of smart warehouses, facilitated by a diverse array of warehouse automation technologies, directly addresses another critical challenge in inventory management: maintaining real-time access to accurate product information within the warehouse. This convergence of AI and Industrial Internet of Things (AIIoT) presents a potent solution, fostering a paradigm shift towards proficient and optimized inventory management practices. This sensor network gathers a comprehensive range of data crucial for informed decision-making. Leveraging AI, the system analyzes parameters like product location, environmental conditions, and equipment performance metrics, enabling intelligent optimization of warehouse operations [17].

Within smart cities, the strategic integration of advanced technologies and interconnected systems fosters a data-driven decision-making environment. This extends to the realm of warehouse operations, enabling them to achieve unprecedented efficiency and precision in inventory management and tasks, ultimately optimizing overall process flows as shown in Fig. 2, empowering businesses to leverage real-time data for enhanced efficiency.

Fig. 2.

The warehouse flows that feed systems.

Warehouse automation encompasses the process of automating inventory movement with minimal human intervention. This automation can be achieved through various means, not necessarily limited to physical robots. In many cases, it simply involves leveraging software solutions to streamline and replace manual tasks associated with inventory management [9]. However, the high costs associated with the commercial systems limit its accessibility, motivating innovative low-cost solutions aimed primarily at automating the WMS, using software and data analysis to optimize operations [4].

The level of automation within a warehouse directly correlates with the standardization of goods movement [9]. Highly homogeneous goods with narrow tolerance levels are ideal for extensive automation. This maximum degree of automation is achieved when a central computer manages all movements through a network of fully automated equipment. This necessitates the implementation of such equipment alongside the programming of comprehensive loading/unloading plans for the storage system [4].

Effective automation initiatives necessitate the synergistic integration of a diverse technological landscape. In this context, an AIIoT solution facilitates the real-time acquisition of data through a network of interconnected sensors, this hybridized approach presents a transformative solution with the potential to revolutionize diverse sectors and unlock unprecedented opportunities for both businesses and individuals.

Therefore, this project assesses the implementation of a reference model based on free software platforms focused on demand forecasting, emphasizing the optimization of the trade-off between inventory management costs and the maintenance of sufficient stock levels to satisfy customer demand and avoid stockout scenarios.

3.1.System design

The proposed solution leverages AIIoT functionalities. The AI component ingests both real-time and historical data from existing WMS. This data is then utilized to dynamically update inventory reports and KPIs. Furthermore, by incorporating a demand forecasting module, this predictive analytics framework expands its capabilities to project future sales requirements and optimal safety stock levels [25]. This approach facilitates the strategic deployment of inventory across designated storage locations within the warehouse facility.

Within the IoT, interoperability ensures seamless communication and data exchange between devices, irrespective of their manufacturer or brand [15]. This capability facilitates the aggregation of data and access to a wider range of functionalities from a unified platform. The underlying protocols significantly impact the overall functionality and efficacy of data transmission across the internet. These protocols, which are often dependent on the specific system architecture, can sometimes incorporate features like network interface level address resolution and auto-configuration [14].

To ensure the successful development and construction of the proposed system, a phased approach is recommended. This methodology entails the strategic subdivision of the project into distinct stages, each with clearly defined deliverables. These stages typically encompass automation configuration, solution delivery, and architectural design. This phased approach fosters improved project management, facilitates risk mitigation, and allows for iterative refinement throughout the development lifecycle.

This optimization prioritizes customer satisfaction by enabling the fulfillment of projected demand through the utilization of predictive analytics.

3.2.Development phases

Optimizing warehouse operational management necessitates a holistic analysis of its constituent sub-processes. These interdependent sub-processes form a critical chain, requiring a thorough evaluation to identify bottlenecks and inefficiencies. This analysis should delve into the specific needs and objectives of the target warehouse. By doing so, it facilitates the identification of improvement opportunities and the subsequent definition of relevant KPIs that directly align with the desired operational outcomes [16].

Different types of warehouses have their own storage allocation mechanisms based on their needs and requirements [10]. To adequately predict product needs and generate ideal inventory policies, the integration of different technologies that allow identification, registration of operations and traceability is required.

The system must allow parameterizing the relevant information for the correct management of the warehouse as a smart unit, presenting a series of reports considering the presenting an overview about specific items and storage units and their availability in inventory [16], an appreciation of inventory valuation that provide historical information on the value of inventory inputs and outputs over time, and the age composition of the selected products in the inventory in terms of quantity and value.

To develop the forecasting algorithm, the value of the coefficient of determination (R2) is considered, which indicates the proportion of the variance of the dependent variable that is explained by the regression model used to predict that variable.

Finally, the inventory forecasting solution will be integrated into the WMS, or the system intended for inventory control. Real-time and remote data exchange will be enabled, empowering improved inventory management practices.

3.3.System architecture

Implementing an automatic inventory control system requires complementing the WMS with a product tracking system in the warehouse. In addition, the aim is to integrate the data into a centralized platform for processing and analysis, since large volumes of data are managed in this core, so an Artificial Neural Network (ANN) is applied for the analysis and generation of reports on the status of the data inventory, allowing, among other actions, optimize inventory levels to meet demand without overstocking setting the reorder point of certain items based on the estimate [22]. The application front end must be compatible with complementary applications that allow inventory to be always tracked as well as with mobile devices.

The proposed architecture for the prototype (Fig. 3) consists of tracking merchandise inventoried in a warehouse using a development platform with a Particle Photon microcontroller unit (MCU), in addition to RFID tags and sensors, mobile devices, with complementary cloud applications.

The demand forecasting algorithm leverages historical data encompassing sales figures, promotional events, and market trend indicators. This comprehensive data analysis enables the generation of accurate forecasts, empowering proactive inventory management decisions. These decisions include the determination of optimal order quantities and product types required to maintain minimum stock levels. Additionally, cloud-based technologies facilitate the accessibility and deployment of a wider range of technological solutions, further augmenting the capabilities of the forecasting algorithm.

Fig. 3.

Warehouse automation’s block diagram.

Inventory data patterns can be detected and provide valuable information on products that have a higher turnover, as well as those whose demand is seasonal, allowing the optimization of storage space and analyzing supply chain data in its entirety joint identifying bottlenecks and other issues that could affect the availability of products in inventory.

In this way, goods into, within, and out of warehouse are detected by the RFID sensor (Fig. 4) and all events are documented within a designated log. This log is then processed by the console, which stores the data in a database. The EOQ method is subsequently employed to analyze the information and generate calculations, obtaining data on the available inventory, which is monitored automatically using the MCU (Fig. 5), until the reorder point – demand during delivery time.

Fig. 4.

Prototype of the RFID sensor connected to the MCU.

Fig. 5.

MCU and RFID reader connection electrical diagram.

The model undergoes a normalization process to establish a consistent demand baseline. Leveraging this normalized data, the AI then predicts future demand with the objective of minimizing total costs across the supply chain [18]. This cost minimization incorporates both inventories holding costs and order preparation/placement costs. Automation within this specific segment of the supply chain can be initiated through the implementation or enhancement of a WMS. By placing a strategic focus on data collection and meticulous inventory control, such a WMS improved is expected to generate significant economic benefits.

When planning to implement such a solution for asset tracking and management in the warehouse, it is essential to consider data security and privacy concerns. Therefore, it is important to take these concerns into account and choose the best security measures; the platforms used allow you to configure firewalls, user authentication and data encryption, to protect the IoT infrastructure.

4.1.Results analysis

To implement the proposed architecture in the WMS, in the first instance modelling was carried out as an information system connected to several internal and external systems, developing a prototype where the different hardware modules interact with the specific software, which allows the integration of an IIoT system.

In the warehouse case study, the data is integrated by detecting RFID tags generating the receipt of goods (Fig. 6), which are added to a test database. The data normalized is stored in a separate repository where it is available and ready for exchange.

Fig. 6.

RFID card reading sequence using a Particle Photon MCU.

A process was implemented to migrate the data from the visualization console through a connection with the database for registration, validating its integrity and consistency. It is necessary to ensure that the transfer and storage process complies with the best security and access control practices in accordance with institutional policies.

The historical data is based on the classification of data from different time ranges that constitutes the ingest for the forecast model. In this way the data stored in the database also can be displayed raw to allow visibility of the data and movement of goods in the warehouse. Subsequently, an algorithm was designed for the forecasting of inventory needs (Fig. 7), which performs preprocessing on the data by dividing it into a training set and a test set, which depends on the size, complexity, and availability of the inventory data.

Fig. 7.

Example of an ANN algorithm for inventory forecasting in Python.

It is recommended to have a training set large enough to capture the variability and important characteristics of the data. In addition, it is also important to have a suitable test set to evaluate the generalization ability of the model after training.

The efficiency of the solution is measured in terms of time and accuracy with the prototype tests in the proposed architecture, it is important to note that a manual forecast of inventory reorder through an EOQ method can be efficient and relatively simple to implement in situations where the demand and costs are constant and predictable, helping minimize the total ordering cost incurred throughout the year. However, in more dynamic and complex environments, they could use other predictive models, in this sense, AI has the advantage of being able to better handle uncertainty and changing patterns to analyze data and predict what will happen to inventory in the immediate future.