2.1. Components

Two specifications of Bluetooth components, including BLE 4.2 and BLE 5.0, are used as beacons and scanners to evaluate the signal stability and effectivity in experiments. As the Bluetooth Special Interest Group (SIG) announced, BLE 5 significantly increased the range, speed, and broadcast messaging capacity compared to the previous BLE 4 [19]. It is still necessary to verify the signal attribute difference in the field between them. Furthermore, since BLE 4 has been developed and used for an extended period, there are many application deployments. Therefore, these two, new and old, have incredible opportunities to work in pairs in the application field.

The RSSI-based indoor positioning system should contain transmitters and receivers, known as a beacon and scanner, respectively. There are two types of configurations for the deploy beacon and receiver: tag-based [20] and beacon-based [21]. This study used the beacon-based configuration, with the beacons placed in a fixed position, and the scanner acted as a moving element.

The selected components for the experiment are both microcontroller units (MCU): Arduino Nano 33 and Linkit 7697. Their main features are shown in Table 2. The selection of MCUs relies on three main factors: standardization, customization, and flexibility.

Specifications of the components (BLE 5.0 and BLE 4.2).

¹ Nordic nRF52840 chipset: https://content.arduino.cc/assets/Nano_BLE_MCU-nRF52840_PS_v1.1.pdf (accessed on 28 July 2021). ² ACSIP wrtnode7 chipset: http://www.acsip.com.tw/index.php?action=technical&p=2 (accessed on 28 July 2021).

Firstly, the multiple offer of Beacons in the market is enormous, from companies such as BlueCats [22], Estimote [23], or Kontak.io [24] providing commercial versions of beacons. To simplify this issue, our approach focused on the use of similar hardware for transmission and reception in order to guarantee a more standard procedure. Secondly, the customization of critical parameters (advertising the interval and scanning times) from emitters and receivers is crucial to generate a significant amount of data, guarantee an appropriate sampling rate, and control the overall system latency. Finally, the selected hardware (Linkit 7697) offers both Wi-Fi and Bluetooth connectivity, and it is able to support the MQTT and CoAP protocols, facilitating the interconnectivity and providing flexibility on the way to achieve this interconnection between the receiver (scanner) and the database.

Based on the technical specifications, the transmitting segment in dBm (decibel relative to one milliwatt) for Linkit 7697 is slightly higher than Arduino Nano. On the other hand, BLE 5.0 can perform a little better sensitivity in the receiving segment: −103 dBm at the long-range mode and −95 dBm at the low-energy mode. To make the experimental data clearer and readable, Linkit 7697 will work as the scanner. The scanners were used to collect the transmitting signal of beacons Arduino nano 33 and Linkit 7697, respectively.

The appearances of those two components are shown in Figure 2a,b: Arduino nano 33, BLE 5.0 and Linkit 7697, BLE 4.2. Both use a ceramic antenna design, so there is no antenna rod in appearance. Generally, the flat antenna makes it more convenient for conducting a test with less consideration toward directional issues. However, in practice, the antenna orientation still causes a deviation of the signals among the devices.

The appearance of Arduino nano 33 and Linkit 7697.

As mentioned previously, both devices use a ceramic antenna for emitting and receiving signals. The ceramic antenna offers several advantages, such as multiple configurations, a small size, less sensitivity to components, and less environmental noise. However, its main shortcoming is the slight lack of performance compared to the PCB trace antenna [25].

Additionally, the antenna can only be on one side of the circuit board. Thus, to understand the influence of antenna orientation on signal reception, two positions were selected for testing, “face-to-face” and “back-to-back”, as shown in Figure 3a,b.

Antenna orientation: (a) face-to-face and (b) not face-to-face.

The orientation influences the RSSI between the beacon and scanner, regardless of their model. We conducted measurements with Arduino nano 33 as a beacon and Linkit 7697 as a scanner to prove this effect. The results are shown in Table 3.

Antenna orientation influence on the RSSI.

¹ RSSI: Received Signal Strength Index, unit: dBm; mean value (1000 samples).

According to the sensitivity of the scanner, as shown in Table 2, signals under −94 dBm cannot be considered as a correct lecture. It is possible to observe that the back-to-back antenna configuration caused a significant impact on the strength of the reception or RSSI even at a short distance, as shown in Table 3. On average, the back-to-back orientation decreases 19 dBm in comparison to the face-to-face antenna orientation. This difference can generate significant location positioning errors. For this reason, all measurements conducted in this study remained in a particular orientation, face-to-face, as mentioned in Figure 3a.

The experiments of this work were performed at the Industry 4.0 Implementation Center from the National Taiwan University of Science and Technology (NTUST). The center is an educational shared demo factory that contains multiple spaces, such as a workshop, tool room, laboratory, meeting room, and classroom.

The main experiment fields include two different environments: a classroom and workshop, as shown in Figure 4. The classroom space is regarded as a relatively low-noise environment, with dimensions of 11.5 m by 6.5 m. On the other hand, the workshop is regarded as a relatively noisy environment, with dimensions of 5 m by 5 m (one-quarter of the space). Besides the size limitation, the workshop space contains heavy machinery around its surroundings, providing a closer approach to real-life factory environment conditions.

Experiment field layout (workshop and classroom).

An internal numerical system was imposed to differentiate the hardware components in both environments. Thus, for the classroom, Arduino nano 33 has numbers #51–56 (6 pieces) and Linkit 7697 has numbers #61–66, respectively. Similarly, Arduino Nano33 has been assigned the numbers #71–79 (9 pieces) and Linkit 7697 #81–89 (9 pieces) at the workshop, as shown in Table 4.

Components numbering (BLE 5.0 and BLE 4.2).

¹ Classroom: less noisy environment (refer to Figure 5c). ² Workshop: general factory noise (refer to Figure 6c).

Positioning systems such as real-time locating systems (RTLS) are time-sensitive and require a stable positioning update interval. Therefore, choosing an appropriate and effective communication architecture is very important to reduce data lagging or call latency.

Following the TCP/IP architecture, it is possible to decompose the whole IPS paradigm into four main layers: link, internet, transport, and application [26]. First, the link layer defines both the physical and data link protocols, such as Bluetooth, 802.11 b/g/n, and MAC. Microcontrollers (MCU) and General Purpose Units (GPU) are in charge of this layer. Second, the internet layer relies on the Internet Protocol (IP) to pack and deliver higher layers. For example, access point (AP) devices support IPs on both versions (IPv4 and IPv6), using ports 1027 and 3784, respectively. Third, the transport layer relies on the Transport Control Protocol (TCP), and routers and firewalls handle TCP packages. Finally, the application layer hosts Message Queuing Telemetry Transport (MQTT), working primarily on TCP/IP, using ports 1883 and 21883, respectively. The Table 5 interconnection model and protocols are conducted in the following experiments.

Interconnection model and protocols.

¹ Open Systems Interconnection model (OSI model), https://en.wikipedia.org/wiki/OSI_model (accessed on 28 July 2021). ² MQTT broker, https://en.wikipedia.org/wiki/MQTT (accessed on 28 July 2021).

Servers and storage components rely on the use of the MQTT protocol for interconnections. The package structure follows a JavaScript Object Notation (JSON) stored in MongoDB, a NoSQL database. Implementation of the Bluetooth stack can vary according to the type of hardware. Therefore, it is possible to have a General Purpose Bluetooth Stack (BlueZ) [27] or Embedded system implementations (BlueMagic) [28]. The selected device used for research on this topic, MT7697, works with a Qualcomm version of BlueMagic (MediaTek) [29].

General-purpose data collection systems monitor the value of the signal and export single values in comma-separated values (CSV) stored locally every 500 ms [30]. Using embedded systems to harvest data with the proposed protocol architecture allows generating over 200,000 samples for two hours or an average of 32 ms per sample. Thus, the suggested framework to collect, pack, transport, and deliver data to the database provides a solid base to develop IPS solutions to generate a significant amount of data in a shorter period, improving the Overall Operation Effectiveness (OOE).