The ultra-high frequency (UHF) band was selected as the most suitable RFID type, especially in cases where energy transfer is crucial for passive RFID distancing. Out of the multiple choices, the option of the ISO-18000-6C standard has been preferred, given its widespread recognition (18). In contrast to the more frequently used LF (low frequency) 125 kHz RFID systems, typically used for close-range ID scanning with a limited reach of less than 30 cm, the UHF option chosen here operates in the 860–960 MHz range, making it more suitable for mid-range areas.

For tracking items dispersed across multiple floors, the UHF passive RFID system is ideal, offering a scannable range of over 1 m and extending up to 15.5 m (19). This method offers a valuable contribution by presenting an affordable and scalable solution, applicable to various fields such as asset tracking and environmental monitoring. It ensures efficient performance while catering to a broad spectrum of use (3, 20).

The range-based method in this study works on the principle that signal strength decreases as the distance between the RFID reader and tag are increasing, which making the RSSI values a good proxy for estimating the distance (21–23). Accurate distance measurement is very important to effective localization of monitoring objects, with RF providing the basic physical property for this purpose. The speed of electromagnetic waves is giving a reliable way for estimate the distance between the transmitter and receiver. Many systems, like radar and sonar, use metrics like Time of Arrival (ToA) and Time Difference of Arrival (TDOA) to decide this spatial relationship (24), which are common in global positioning systems and light detection and ranging technology (25). ToA finds the distance by measuring the travel time of a signal between transmitter and receiver, then multiplies this time (T) by the speed of light (C) to get the total distance. This calculation can also be represented as:

representing the range (R) of radar detection.

Alternatively, the decay of an electromagnetic signal can be used to determine distance, as outlined by several parameters calculated in Equation (2).

Where P_R: reflected power received [W], θ: the gain penalty, B: the path loss due to blockage, P_T: reader transmitted power [W], G_t: tag antenna gain, G_T: transmitter antenna gain, λ: signal wavelength [m], X: polarization mismatch, F₂: the fade margin for monostatic, and M: modulation factor (26).

An experiment was carried out to investigate how RSSI relates to the distance between 3 distinct readers and a single tag. The findings indicated a non-linear correlation and data uncertainties, particularly pronounced when the tag was located 50 cm below the bottom surface of ceiling-mounted readers, as illustrated in Figure 1.

Regarding uncertainty, elements like RF polarization (27) or interactions of RF waves can lead to RSSI fluctuations during distance measurement. These variations accumulate errors during triangulation calculations (28), ultimately reducing the accuracy of RSSI-based localization. Several enhancement methods have been proposed to improve accuracy. This includes employing semi-supervised learning to monitor shifts in signal distribution (29), utilizing trilateration that incorporates multiple measurements via Wi-Fi, and round trip time (RTT) in conjunction with the PoseNet categorization learning approach (30), or using device-assisted methods for motion tracking (31). Despite these advancements in location accuracy, including techniques like triangulation (8) and phase analysis (7), challenges remain in achieving precise measurements consistently.

With the increasing computational power of digital microchips, electrical transponders have significantly decreased in size and cost, allowing them to be integrated into consumer-grade devices. These technological advancements have enabled researchers to investigate the distribution of electromagnetic fields across a network of transmitters and transponders. Looking at the LANDMARC system (32, 33), it pioneered the concept of reference tag fields, significantly improving the accuracy of indoor positioning. Through practice, the K-Nearest Neighbors (KNN) algorithm, researchers have developed a path loss model which effectively minimizes the errors (34). In addition, using the nonlinear backscattering method, it is helping passive devices to achieve high-precision instant 3D indoor positioning, with a median error is only 3.5 cm (35). Together, these innovations are representing a big advancement in the localization technology and underscore the potential to more precisely track in an indoor environment.

To overcome the challenge of using RSSI for distance-based localization, this study is adopting a range-independent approach and focusing on determining whether a tagged object is lying within the coverage of a specific reader in the room. By shifting the emphasis from distance measurement to RF sensing boundary, it ensures the reliable detecting of tags in the designated area. To solve the interference which caused by multipath propagation, the direct path RSSI is prioritized as the primary signal, enabling identifying of reader that capture the strongest signal and facilitate accurate room-level localization without needing complex calculations. Additionally, RSSI statistical analyses are used to define a virtual room boundary through a change point, simplifying the localization process and improving both speed and efficiency. This comprehensive strategy is providing a practical and effective solution for ensuring reliable indoor localization.

The experiment involved the use of ISO18000-6C EPC Class 1 Gen 2 passive RFID tags, which operate within the 860–960 MHz frequency range. These RFID tags were chosen for their affordability, priced below a quarter of a U.S. dollar each. Their compact size, measuring 73.5 mm by 21.2 mm with a microchip thickness under 1 mm, makes them highly portable and popular for widespread applications. Each tag comprises 3 core elements: a microchip, an inductive circuit, and a dipole-end load. As passive devices, they rely on the energy from the reader’s signal to emit an ID signal intermittently. The reader’s antenna charges the electricity capacitor and powers the microchip to facilitate this process.

For the experiments, an Alien ALR-9650 UHF reader was selected due to its compatibility with the RFID tags. This reader features a polarized antenna and houses an integrated microcontroller within a compact unit of 23.2 cm × 22.9 cm × 5.1 cm, making it ideal for ceiling installations. Additionally, the reader supports Power over Ethernet (PoE), allowing it to both receive power and transmit data via a single cable, improving its efficiency and versatility in various environments.

In standing RFID systems, the reader faces a bandwidth limitation, which hinders its ability to simultaneously detect multiple transponder signals. When multiple data signals overlap within the same bandwidth, the reader becomes confused, making it hard to differentiate between signals that start simultaneously. To solve this issue, an anti-collision mechanism is used, which is processed by readers before detection (36). A commonly employed anti-collision method is the binary tree-based slotted ALOHA protocol. In this approach, time is segmented into discrete slots, with each tag receiving a unique identification code (37). With this algorithm, the reader is dividing the time into the time slot and giving a unique identification number (ID) for each tag, organizing the identification process more efficiently by using the binary tree structure.

The slotted ALOHA algorithm, which is a method for efficiently identify RFID tags, can be mathematically represent as follows: Given a system with N transponders, each has a unique identifier, and the algorithm proceeds in 3 steps:

The reader broadcasts an initial query to all tags, signaling them to respond in the first time slot.

After receiving the query, each tag is randomly delaying its response and then transmitting its unique identifier in the assigned time slot.

If multiple tags respond in the same time slot, a collision has occurred. To resolve this, the reader uses a binary tree structure to identify each colliding tag individually.

The slotted ALOHA algorithm, which is a clever technique using binary trees, effectively minimizes the conflict and ensures the unique identification of every tag’s response. To enhance system performance, we devised an anti-interference algorithm inspired by solutions for wireless communication challenges. This algorithm is addressing the potential interference which is caused by tag signals in the RFID system.

At the key of this solution is a duplicate ID filter. The filter is monitoring the incoming tag responses, storing their IDs and the RSSI value in the buffer. Then, the moving average calculation is applied on the RSSI values for every ID, which provides a reliable reference for future comparison.

When a new tag response has arrived, the system is cross-referencing its ID with an entry in the temporary buffer. If a match is found, it signals a potential duplicate ID, which indicates the possible interference. To solve this issue, the algorithm is comparing moving average RSSI values of tags that have duplicate IDs. The outliers in those values, which are often caused by weak signals or interference, are identified and excluded from the moving average calculation. This filtering process not only enhances the accuracy of interference detection but also significantly reduces computational overhead.

RSSI is playing a crucial role for identifying the transition point between RFID readers and establishing the virtual boundaries to demarcate the adjacent area. When an RFID tag is transitioned from one reader’s coverage to another, a decrease in RSSI is observed on the former reader, while a corresponding increase is noticed on the latter one. This change point is serving as a virtual boundary, distinctly separating the different zones or rooms.

Figure 3 illustrates the application of the change point algorithm in an RFID setting. In this scenario, an RFID tag, symbolized by an arrow, moves at a constant speed of 4 cm/s between 2 RFID readers. The tag’s journey commences at time 0 and spans 60 s, with the tag positioned 100 cm above reader 1. As the tag moves towards reader 2, the system records a gradual decline in RSSI on reader 1, accompanied by a simultaneous increase on reader 2. This pattern of RSSI fluctuations clearly delineates the virtual boundary between the 2 readers.

At approximately the 34-s mark, Figure 4 reveals a critical intersection where the signal strengths of the 2 readers converge, suggesting a possible change in the tag’s position. Notably, from 30 to 45 s, both readers show a consistent decline in RSSI readings. This decline in signal strength can be attributed to several factors:

When RF waves interact with obstacles like metal poles or building corners, they undergo reflections and diffractions, leading to multipath effects. These effects cause the signal to travel along multiple, often indirect, paths. The indirect paths disrupt the direct communication link between the RFID tag and reader, causing phase differences that result in interference. This interference can lead to signal attenuation or cancellation, creating “zeros” in certain areas of the read field, ultimately lowering the RSSI.

Materials such as hidden wooden or metal strips within walls can reflect or absorb RF waves, weakening the signals. This interaction causes the signals received by both readers to be diminished as the energy of the waves is either redirected or absorbed by the surrounding environment.

Given the complexity of electromagnetic wave behavior and its sensitivity to environmental factors, which can introduce uncertainty in distance measurements, we designed an anti-interference algorithm. This algorithm identifies the strongest RSSI signals between the RFID reader and tag, ensuring more accurate distance calculations. Furthermore, the effectiveness of using RSSI for detecting the location of objects in nearby areas was confirmed through the RSSI change point detection test.

The data acquisition system was developed using NI LabVIEW, a graphical programming tool that enables seamless communication with the RFID reader through assigned IP addresses and ports. The libraries in LabVIEW include Internet modules designed specifically for this purpose. To ensure accurate data recording, we implemented an instant pooling algorithm that centrally timestamps each detected RFID tag. The collected data is then combined with the location of each reader and forwarded to the low-congestion queuing processor.

In the software, each reader and writer is assigned a dedicated queue, which is subsequently merged into a main pool for anti-interference via a pooling mechanism. Figure 5 demonstrates the flow of data for an RFID tag, showing how it follows the anti-collision protocol. Within a dynamic interval of under 1 s, the reader sends a packet, including the tag ID and RSSI value, to the server over the Ethernet network. Each reader operates autonomously, transmitting its packets independently. Upon receiving the data, the server designates a port and module for each incoming connection, corresponding to the specific area being monitored by the RFID system.

To enhance the efficiency of tag conflict resolution, the task was initially assigned to each reader’s computer, rather than relying solely on the server’s central processor. The next step involves the server employing multi-threading modules that operate in parallel, handling data from several incoming ports simultaneously. When data is received at a specified port, the module identifies the corresponding ID and RSSI set, storing them in designated cache slots. During this process, centralized timestamps are applied to each ID set.

These cache slots serve multiple functions: filtering out duplicate or corrupted data, calculating the moving average of RSSI values, and synchronizing the queues. Each module operates independently, with its own clock, adhering to the first-in-first-out (FIFO) protocol. Once the data is processed, all modules compile their ID sets into queues for the anti-jamming process. A major challenge in this system is interference, which is addressed in the Methods section. After dequeuing the sparse data into a central counter-shading pool, we apply a change-point algorithm to pinpoint the ID set’s location by refining its RSSI data.

The processed and aggregated tag data is stored in a web-based storage system to facilitate real-time visualization of each tag’s location. A Python script can then retrieve this data and plot the tag locations on a floor map. The visualization can include additional details such as the tag’s ID, ID group, free time, icon design, and other relevant information. Furthermore, system terminals can access the aggregated data via the cloud, enabling quick retrieval of information from anywhere in the world.

A field trial was conducted to track the movement of coffee cups equipped with RFID tags. Office staff participated by carrying labeled cups for a designated period of time, allowing us to evaluate the system’s performance in a real-world setting compared to the original laboratory setup. After repeated experiments to improve the system, we successfully tracked the tagged coffee cups, as shown in Figure 6. The coffee cup’s location was accurately tracked, demonstrating the system’s effectiveness in a real-world environment.

A bird’s-eye view of a typical office, with 3 interconnected rooms, is visualized on the front-end interface of the RFID system. To avoid confusion during interruptions, only 1 room’s label markers are displayed at a time. The anti-interference system effectively prevents multiple readers from detecting the same tag signal, ensuring accurate tag position identification. Distinctive icons, varying in color and shape, represent different states. These icons can also incorporate label names, object group symbols, and color codes to signify specific durations. As depicted in Figures 7a and b, fluctuations in the number of tracked objects, their locations, and the duration of room occupancy over a specific time range are illustrated through updates.

A dashboard diagram is illustrated in Figure 8, where each block corresponds to an object. A table listing registered tags is positioned next to the yellow bar. Upon receiving a tag signal, the system determines the tag’s location and initiates the tracking of its dwell time, represented by the length of the yellow bar. This dashboard facilitates accurate monitoring of room occupancy.

During the recording process, historical tag data is transformed into bitmap footprints, allowing for the tracking of each object’s movement. The real-time dashboard then displays the current status of these objects. Figure 9 shows the raw data for each tag, with RSSI readings represented as grayscale squares; lighter shading corresponds to stronger and closer signals. A 3 × 100 grayscale bitmap forms a long rectangular box. The horizontal axis represents 10-s intervals, while the vertical axis is divided into 3 rows, each corresponding to one of the 3 reading rooms. This dashboard uses 100-bit squares to visually illustrate how the reader detects tag signals over a 500-s period.

Figure 10a graphically illustrates the interference rejection process. After 250 s, the label transitions from Room 2 to Room 3, resulting in a distinct original transition in the center of the bitmap. Although Room 3 showed stronger RSSI readings after the conversion, Room 1 continued to record weaker RSSI signatures along the path. To solve this problem, the anti-interference system identifies the strongest signal among all readers, identifying Room 3 as the most likely location of the tag, as shown in Figure 10b.

Finally, Figure 11 shows the footprint dashboard showing all tracked objects, designed to ensure that each tag is only visible in 1 room at any given time. For example, Track 1 shows the movement of an object from Room 2 to Room 3, while Track 4 shows the transition of an object from Room 1 to Room 3 and did not return to Room 1.