The Internet of Things (IoT) is a network of internet-connected devices that contain embedded sensors and actuators to communicate with the external environment [1
]. Because of its widespread applications in various areas, such as in healthcare, agriculture, sports and security, the number of internet-connected devices is increasing at a rapid rate [2
]. The IoT market is growing continuously, and it is expected that, by 2025, global IoT market revenue will rise to around 1.6 trillion U.S. dollars (USD) [4
]. The IoT-enabled wearable sensor devices have become a promising technology that enables applications such as continuous wireless monitoring of vital physiological parameters such as arterial oxygen saturation (SpO2
) and heart rate (HR) [5
]. Among the various wearable devices available, it is forecasted that the end-user spending on smartwatches could exceed 27 billion USD in 2021 [6
Photoplethysmography (PPG) is a technology that uses optical sensors for non-invasive monitoring of the volume changes in arterial blood [7
]. PPG is widely used in pulse oximetry for the measurement of various health parameters such as SpO2
, HR, blood pressure, anaesthetic depth and respiration rate [9
]. Some companies, such as Apple, Huawei, Fitbit and Fossil have recently added SpO2
monitoring features to their wrist-worn devices [11
]. Estimating the oxygen saturation level in the bloodstream helps in determining the sleep pattern and the oxygen saturation status of a person during sports, fitness, and many other physical activities.
Wireless communication is a key technology in continuous sensing and monitoring the health conditions of a person [15
]. Generally, the sensing device has a limited computation and processing capability, and typically, an additional processing unit is incorporated for data processing [16
]. Further, a series of operations such as data visualization for the end-user, and transmitting the measured data to other centers can be accomplished using a smartphone application [16
]. Typically, wearable healthcare sensors and systems require constant accessibility and operation of the sensed data [16
]. Advancement in cloud computing technology offers a great opportunity to transmit the sensor data to the IoT cloud, and when required, an authorized user can easily access the data with any internet-connected device [19
Currently, the majority of wrist-worn devices operate in the conventional 2.45 GHz industrial, scientific and medical (ISM) frequency band [12
]. Wireless communications at the 2.45 GHz band have several advantages, such as worldwide availability, a higher data rate, compatibility with a large number of wireless standards, and small size antenna [5
]. However, the 2.45 GHz band has now become highly crowded, which can dampen the communication reliability and the quality of service (QoS). [5
An emerging alternative to the 2.45 GHz ISM band for IoT applications is the Sub-GHz band [5
]. The most popular Sub-GHz bands include 433 MHz (Asia), 868 MHz (Europe), and 915 MHz (USA) [28
]. In [31
], the development of a wearable tracking system using the Sigfox protocol is reported, where the implemented 868 MHz meandered inverted-F antenna requires an area of 270 × 75 mm2
. A wearable long-range (LoRa) sensor device for tracking and wireless health monitoring is reported in [32
]. However, in this study, the implemented hardware components were not developed, and instead, commercial radio and whip antenna were used. In [33
], a 915 MHz LoRa-based wearable sensor platform for IoT health monitoring is reported, where a commercial antenna was implemented, and the height of the developed device is large (45 mm). In [5
], the authors developed a 915 MHz wristwatch-based wireless sensor device, where a comparison of the communication performance at 915 MHz and 2.45 GHz is also discussed; however, the developed sensor device was not fully functional. To summarize, the majority of the wearable wireless sensor platforms reported in the literature use a commercial radio and antenna and require a large size for their implementation [31
A number of Sub-GHz band wrist-worn antennas have been reported in the literature [5
]. An 868 MHz helical wristband antenna reported in [35
] has a −10 dB impedance bandwidth of 17 MHz but has a low peak realized gain of −13 dBi. A dual-band wrist-worn loop antenna reported in [36
] has a −10 dB impedance bandwidth of 180 MHz; however, it has a low radiation efficiency of 6.4%. Furthermore, a wrist-worn dipole antenna reported in [37
] has a wide impedance bandwidth of 160 MHz; however, this antenna has a low radiation efficiency of 6.3% at 900 MHz. The authors present a 915 MHz wristwatch planar inverted-F antenna (PIFA) in [40
]. This antenna has an on-body measured −10 dB impedance bandwidth of 26.4 MHz, a peak realized gain of −0.57 dBi and a radiation efficiency of 46.8% at 915 MHz. However, the height of the antenna from the nearest ground plane is large (10.3 mm) when compared with other designs reported in the literature [5
]. In [5
], the authors present a 915 MHz meandered PIFA wristwatch-integrated antenna. This antenna shows an on-body peak realized gain of −6.1 dBi and a −10 dB impedance bandwidth of 55 MHz. The majority of Sub-GHz band wrist-worn antennas reported in the literature do not exhibit satisfactory impedance and radiation performance simultaneously [35
]. In addition, most of the reported antennas (except [5
]) assume an ideal environment and do not take into account the effects of the nearby internal components of the wristwatch device.
The key focus of this work is to develop a compact wireless sub-system for 868 MHz and a gateway to establish wireless communication between the wristwatch sensor platform and a smartphone. The novel aspect of this paper includes the design of a compact on-body 868 MHz antenna and a radio that fit into the developed wristwatch-based wireless sensor platform. In contrast to other research efforts reported in the literature, the developed 868 MHz antenna includes all the internal components of the practical wristwatch sensor device and focuses on achieving improved impedance and radiation characteristics. In addition, this work also evaluates the wireless communication range performance of the developed wristwatch sensor platform in an indoor environment and demonstrates a greater communication range when compared with a commercial Bluetooth (BLE) wristwatch device.
The remainder of the paper is organized as follows: a comparison of wireless performance in the 868 MHz and 2.45 GHz ISM bands is discussed in Section 2
. The system design aspect of the developed sensor platform is described in Section 3
. In Section 4
, the antenna design, fabrication process, and measured results are described. The communication range measurements of the developed sensor platform are presented in Section 5
. Section 6
presents the sensor platform implementations and clinical trial settings. Finally, in Section 7
, the main conclusions and future research plans are summarized.
4. Antenna Design
This section presents the design, fabrication and measurement of a compact 868 MHz wristwatch-integrated antenna. The design of a compact sub-GHz antenna to fit into a small area contained within a wristwatch device is challenging due to size constraints. In this case, the challenge is to integrate an antenna with a guided electrical length of 57 mm or approximately λg
/4 in a wristwatch device. The key performance parameters of the antenna, such as gain, efficiency, and bandwidth are fundamentally limited by the antenna size [59
]. The geometry of the proposed antenna is shown in Figure 6
. The antenna has a modified planar inverted-F (PIFA) topology. The antenna is excited at point S
using an unbalanced feed with respect to the ground signal at point G
. The signal is applied from the radio board using two contact springs at points S
. The antenna topology is comprised of two resonators, R1
, for improving the radiation performance of the antenna. The particular shapes of R1
were designed in order to keep the resonators away from the internal conducting components of the wristwatch device. An inductive shunt section between points S
has been used to improve the impedance matching of the antenna by compensating the capacitance of the resonating arms R1
that are in close proximity to the ground plane.
The overall antenna structure is printed on the surface of a wristwatch enclosure using the LDS method. The wristwatch enclosure is made of an acrylonitrile styrene acrylate (ASA) material, which has a measured relative permittivity of εr
= 2.9 and a loss tangent of tanδ
= 0.033 at 868 MHz. The electrical properties of all the dielectric materials were measured using a dielectric assessment kit (DAK) from Speag [60
]. The maximum dimension of the antenna is 35.2 × 33.4 mm, and the total dimension of the wristwatch enclosure case is 53.9 × 44.8 × 17 mm.
a shows the exploded view of the wristwatch case model used in the electromagnetic (EM) simulations. The wristwatch case includes all the internal components of the device, such as a sensor unit, processing unit, radio board, and connectors. For the demonstrated system, the wristwatch casing is a modified, off-the-shelf assembly and is IP-65 rated [61
], and contains an integrated rubber gasket between the bottom and bottom enclosure assemblies. The rubber gasket ensures a waterproof seal between the top and bottom enclosure. The connectors J1
join the processing unit to the sensor unit and the radio board. The contact springs denoted as S
on the radio board, as shown in Figure 7
a, connect respectively to the points S
on the antenna structure, as shown in Figure 6
. The simulation model, where the wristwatch device is placed on a commercial SHO-GFPC-V1 phantom arm from Speag [62
], is shown in Figure 7
The phantom arm model is comprised of two layers, an outer silicone layer, and an inner skeletal layer. The outer silicone layer has εr = 30 and an electrical conductivity of σ = 0.7 S/m, and a mass density of 1200 kg/m3 at 868 MHz. The inner skeletal layer has εr = 30 and σ = 2.5 S/m at 868 MHz.
4.1. Antenna Simulations
In this section, the simulation results of the proposed antenna are summarized. The antenna was simulated using ANSYS full-wave high frequency structure simulator (HFSS) [63
]. A parametric analysis of the length of the resonators R1
was performed in order to optimize the antenna for 868 MHz operations. It can be seen from Figure 6
that the parameter L1
determines the total length of the resonator R1
, and L2
determines the total length of the resonator R2
. It is shown in Figure 6
that the resonator R1
extends between points S and A, and has a total length (LSA
) of 57 mm. This value of LSA
corresponds to a guided electrical length of 0.28 λg
, which is close to λg
/4 at 868 MHz.
The resonator R2
has a total length of LBC
= 60 mm, which corresponds to an estimated guided electrical length of 0.29 λg
at 868 MHz. The guided wavelength was calculated using the relative permittivity (εr
= 2.9) of the wristwatch enclosure, which is in close proximity to the resonators. As shown in Figure 8
, the value of L1
was varied between 3 to 6 mm, and the value of L2
was varied between 6 to 9 mm. The parameters L1
mainly control the resonant frequency of the antenna, and as expected, increasing the values of either L1
, the resonant frequency decreases continuously.
The parameters L1
also have effects on the impedance-matching of the antenna. However, in comparison to L1
has more influence on the impedance matching. This is expected because the resonance around the 868 MHz band is contributed by the resonator R2
. Thus, by adjusting the values of the parameters L1
, the antenna can be tuned at the desired resonant frequency, and the impedance matching can be controlled to some extent. The antenna is resonant at 868 MHz for L1
= 4 mm and L2
= 7 mm. The final parameters of the proposed antenna are summarized in Table 3
In Figure 9
, the simulated specific absorption rate (SAR) of the proposed wrist-worn antenna on a Speag phantom arm at 868 MHz is shown [62
]. The simulation environment for the SAR evaluation of the antenna is illustrated in Figure 7
b. The peak SAR point on the phantom arm is located near the feed point of the antenna. For an input power of 1 mW, a peak SAR of less than 0.003 W/kg was demonstrated. This figure is less than 0.1% of the maximum permissible SAR limit of 4 W/kg averaged over 10 grams of wrist tissue [64
4.2. Impedance Matching and Bandwidth Enhancement
The input impedance (ZIN
) of an antenna is a key design parameter that determines the reflection coefficient and the impedance bandwidth of the antenna [65
]. In this section, an impedance-matching network to reduce the antenna reflection coefficient while also improving the impedance bandwidth of the antenna is presented. The matching circuit shown in Figure 10
ensures the maximum power transfer from the source (A
) to the antenna and maintains the desired performance even under small detuning effects [66
]. First, a vector network analyzer [67
] was used to measure ZIN
, which was later used to design an optimal matching network for the antenna.
As shown in Figure 10
, a π-type matching network between points B
has been implemented. The matching components include a shunt capacitor C1
on the source side, a series inductor L1
, and a shunt capacitor C2
towards the load side.
The π-type matching network matches the ZIN
of the antenna to a 50 Ω SMA connector at point A
, as shown. The values of the matching network components were optimized using AWR Microwave Office [68
]. The matching network optimization was performed to reduce the power reflection at the input port of the antenna by varying the values of the circuit components C1
. The final realized values of the matching components are listed in Table 4
4.3. Antenna Prototype Fabrication
The proposed antenna was fabricated using an LPKF laser direct structuring (LDS) Protolaser 3D System, available within the Tyndall Microsystems Packaging Laboratory [71
]. For antenna fabrication, the LDS technology was chosen because of its advantages such as low cost, easy integration to the enclosure structure, and no need for additional flex PCB [72
]. The fabrication of the proposed antenna prototype using the LDS technology was completed in the following steps:
3D part fabrication or selection: The 3D part/object on which the metal has to be printed can be fabricated using standard 3D printers. The commercially available thermoplastic parts, such as metals, plastics, glass, FR4, can also be used. In this work, the antenna structure is printed on an ASA thermoplastic wristwatch enclosure from OKW enclosures [61
Part coating with ProtoPaint epoxy: The part is covered with the LPKF ProtoPaint LDS epoxy layer;
Laser direct structuring: The LPKF laser system creates an outline of the conductive pattern of the design. In this step, the laser removes some of the epoxy material and forms a rough surface on which the copper can firmly adhere during metallization;
Metallization: This step involves the electroless copper plating of the region exposed by laser etching. The photograph of the antenna track during the metallization process is shown in Figure 11
. The metallization of the antenna track was completed in the following four steps:
Step 1: In order to get a low resistance electrical continuity through the activated track on the plastic, copper electroless deposition of the surface was required to make it possible to electroplate it. Using an in-house developed, dimethylamine borane (DMAB)-based copper electroless deposition solution, the track was metallized with copper. The sample was immersed in the bath for 60 min at 70 °C, pH9;
Step 2: The electroless copper deposited on the sample needed to be electroplated up with copper. A Schlotter commercial copper bright bath, ACG8, was utilized for this process. The sample was plated for 60 min, 2 A/dm2 at room temperature;
Step 3: Utilizing an in-house developed, low stress nickel-sulphamate-based electroplating bath, the sample was plated for 10 min, 3 A/dm2, at 60 °C. The minimum thickness (tmin) of the electroplated copper is 14.8 μm, as shown;
Step 4: To avoid oxidation of the nickel surface, a commercially available gold Ormex immersion solution by Engelhard was used to finish the surface with gold. The thickness of the gold finish is less than 0.11 μm. This process took 7 minutes at a temperature of 85 °C.
The proposed antenna was fabricated using the abovementioned fabrication steps, and the final developed antenna prototype is shown in Figure 12
4.4. Antenna Measurements
In this section, the measured impedance and radiation characteristics of the wristwatch-integrated antenna are described. In Figure 13
, a photograph of the wristwatch model with an integrated sensor unit, radio board, processing board with an attached battery, and the prototype antenna is illustrated. The Li-ion rechargeable battery used in this work has a voltage level of 3.7 V and a current level of 190 mAh [74
]. For clear visibility, a 3D stack-up of the wristwatch device has already been illustrated in Figure 7
The measurements of the wristwatch-integrated antenna were performed after placing the wristwatch device on a commercial SHO-GFPC-V1 phantom arm [62
]. First, the impedance characteristic of the antenna was measured using a Rohde and Schwarz ZVRE vector network analyzer (VNA) [67
]. In Figure 14
, the measured S11
response of the antenna is illustrated. As discussed earlier, in order to improve the impedance matching and the impedance bandwidth of the antenna, a π-type matching network was implemented. The matching network improves the |S11
| and enables a wideband-matched response around the 868 MHz band. The measured prototype antenna has a −10 dB impedance bandwidth of 36 MHz. This figure is five times higher than the minimum required bandwidth specification of 7 MHz (863–870 MHz) at the 868 MHz band [75
Secondly, the 3D radiation characteristics of the wristwatch device under test (DUT) was measured in an AMS-8050 antenna measurement system [76
]. Figure 15
illustrates the anechoic chamber measurement setup of the DUT. The rotation of the DUT in the chamber is controlled using the multi-axis positioning system (MAPS) or simply the antenna positioner, as shown. The wristwatch device has been placed on an SHO-GFPC-V1 phantom arm [62
] and is supported on a mounting fixture, as shown.
The antenna is excited using an unbalanced 50 Ω coaxial cable. As the connecting cable can affect the antenna properties, multiple ferrite beads have been incorporated in the measurement setup to minimize the effect of the cables on the antenna properties. To illustrate the reliability of the measurement setup, a 2D radiation pattern of a reference 915 MHz antenna that was characterized in a similar chamber setup is shown in Supplementary Figure S1
, where a close agreement between simulated and measured results is demonstrated.
In Figure 16
a, the measured 3D realized gain pattern of the developed antenna at 868 MHz is shown. The coordinate system of the wristwatch device, placed on a phantom arm, is also shown in Figure 16
b. In the xy
-plane, the antenna is characterized as a typical dipole-like radiation characteristic with an omnidirectional radiation pattern. The antenna exhibits nulls along the z
-axis in the yz
-planes. These radiation nulls are expected due to the shielding effect of the finite PIFA ground plane and the absorption of the radiated energy by the phantom arm, along the z
The on-body prototype antenna exhibits a measured peak realized gain of −4.86 dBi and a radiation efficiency of 14.53 % at 868 MHz. To summarize the antenna design aspects, contrary to the Sub-GHz band wrist-worn antennas reported in the literature [35
], the antenna designed in this work includes all the internal components within a practical wristwatch wireless device and the measured results exhibit practically acceptable impedance and radiation characteristics at 868 MHz.
5. Communication Range Measurements of the Sensor Platform
To evaluate the performance of the developed 868 MHz wireless sensor platform, communication range measurements were conducted in an indoor environment, as shown in Figure 17
. In addition, for the purpose of comparison, the communication range measurement for a commercial BLE wristwatch device is also performed [77
]. The indoor environment represents a typical office structure and includes several clutters between the transmitter and the receiver, such as thick and old concrete walls, wooden cubicles, PCSs, tables, chairs and closets.
a shows the front view of the office building where the indoor range measurements were performed. The measurements were taken on the second floor of the building, and for clarity, the 2D map of the floor is illustrated in Figure 17
b. The measurements were conducted after fixing the gateway at point A
, and the position of the wristwatch sensor platform was changed in steps inside the building. In Figure 17
, the point C
represents the edge of the communication at 868 MHz. It is important to mention that the walls between the transmitter (wristwatch sensor platform) and the receiver (gateway) act as an attenuator and negatively affect the wireless communication range.
Typically, in the normal mode, the AT86RF212B transceiver transmits at a power level of 5 dBm [46
]. For the communication range measurements of the 868 MHz sensor platform, the power level of both the transmitter and the receiver was set at 5 dBm and the receiver sensitivity at −110 dBm. For the given indoor scenario, a maximum communication range between points A
and C of approximately 31 m was demonstrated.
In this work, a sensor platform operating at only 868 MHz was developed, and for the purpose of comparison, the BLE range of a Fitbit Charge 3 smartwatch in the same indoor environment was evaluated [77
]. Point B
in Figure 17
represents the communication edge of the BLE Fitbit Charge 3 smartwatch, which is approximately 7 m from point A
Thus, from the range measurements of the developed sensor platform, and its comparison with the Fitbit Charge 3 BLE, it is observed that an improved communication range is possible at 868 MHz. This study shows that when compared to the smartwatch BLE device, communications at 868 MHz offers more than quadruple communication range and is more suitable for indoor applications, such as in offices and hospitals.
6. System Implementation and Clinical Trials
In this section, the implementation of the developed wristwatch wireless sensor platform and the clinical trial results are summarized. The clinical trials were performed at Mercy University Hospital, Cork, Ireland [78
]. “The patients gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of University College Cork, Cork, Ireland, and has been approved by the Clinical Research Ethics Committee (ECM 4 (a) 07/05/19 and ECM 3 (IIIII) 28/06/19)”. The trials using the developed sensor platform were successfully conducted on 24 patients over a period of four weeks, and under the supervision of clinical staff. Each trial takes approximately 3 minutes to measure the SpO2
and heart rate of a patient.
To access the SpO2
and heart rate measurement data on a smartphone application, a gateway was developed, as illustrated in Supplementary Figure S2
. The gateway enclosure has a dimension of 150 × 100 × 45 mm and is made of acrylonitrile butadiene styrene (ABS) material [79
]. The internal components of the gateway include a Microchip SAM R30 Xplained Pro hardware platform [55
], an 868 MHz whip antenna, a 2.45 GHz (Bluetooth low energy) BLE module [56
], two types of LEDs [80
], a 5 V USB power port, and some cables for internal connections. The 868 MHz wristwatch-integrated radio communicates wirelessly with the whip antenna. Initially, both the LEDs (green and blue) remain OFF. When the gateway is powered up by a 5 V supply, the green LED turns ON, indicating that the gateway is ready for the communication. When the blue LED starts blinking, it shows that a successful connection has been established. The nRF52 BLE module is connected to the SAM R30 platform, which has an integrated 2.45 GHz chip antenna [56
]. The BLE antenna transmits the sensor data, which can be received by a smartphone app. The sensor data on the smartphone app could be placed on the IoT-cloud and, when required, an authorized user can easily access the sensor data using any phone-supported cellular network.
In Figure 18
, the devices used in the clinical trials are shown, such as a gateway, an optical biosensor, a smartphone, and the developed wristwatch sensor platform are illustrated. The gateway works as a communication link between the wristwatch sensor platform and the smartphone app. As shown, the clinical trials also employ an existing optical biosensor device, which can wirelessly connect to the smartphone app. Figure 19
illustrates the SpO2
and heart rate measurement setup on a patient’s arm using the developed wristwatch sensor device as well as the existing optical biosensor in a clinical setting. The test protocol used in the clinical trials specified that the arm be parallel to the ground, as the patient would usually be sitting or lying down. This position is recommended because the PPG devices are heavily impacted by motion artefacts, which are likely to have a significant influence on the quality of the measured data [82
The sensor data measured by the wristwatch device were compared against the optical biosensor data. Figure 20
a shows the smartphone app in a user’s hand, showing that both the sensor devices (wristwatch sensor and optical biosensor) are connected. The battery status of the wristwatch and the optical biosensor device can be seen on the app. When both the sensors are placed on the patient’s arm (as shown in Figure 19
) and the patient ID is set using QR codes, the setup is ready for the measurements. The plots of the measured SpO2
and HR sensor data on the smartphone app are shown in Figure 20
b. The green curves represent the SpO2
- and HR-measured values by the wristwatch device, and the yellow curves represent the optical biosensor-measured data. It can be seen that the measurement results from the developed sensor wristwatch device are in close agreement with the results from the optical biosensor device.
This paper has presented the design and development of a novel wristwatch-based wireless sensor platform operating at the Sub-GHz (868 MHz) ISM band, with an example application demonstrated for arterial oxygen saturation (SpO2) and heart rate measurements. Based on numerical calculations and the results reported in the literature, it was shown that the integration of the 868 MHz transceiver offers several advantages such as less path loss, reduced wireless co-existence issues, less attenuation through materials, and improved wireless communication range. However, in contrast to 868 MHz, the 2.45 GHz ISM band is available worldwide, and at the higher frequency, a relatively compact antenna solution is possible. The working principle of a wrist-worn non-invasive optical photoplethysmography (PPG) sensor is also outlined.
The hardware and the wireless software architecture of the sensor platform are proposed. The sensor device incorporates a SpO2 and heart rate sensor and communicates with a gateway using an 868 MHz MiWi network protocol. The MiWi wireless network protocol was implemented because of its potential to offer a low data rate, low power, and low complexity. The implemented radio transceiver uses BPSK modulation, operating with a data rate of 20 kbps, which is more than sufficient for the target application.
The main contribution of this work is the development of a highly integrated 868 MHz wristwatch radio and antenna. The proposed antenna topology is a variant of a planar inverted-F antenna structure and is printed on the 3D surface of a wristwatch enclosure using laser direct structuring (LDS) technology. To improve the impedance matching and to enhance the impedance bandwidth, a π-type matching network was implemented.
The measured on-body antenna exhibits a −10 dB impedance bandwidth of 36 MHz. In addition, the on-body antenna results show a peak realized gain of −4.86 dBi and a radiation efficiency of 14.53% at 868 MHz. Moreover, for 1 mW of input power, a simulated specific absorption rate (SAR) value of 0.003 W/kg at 868 MHz is demonstrated. This figure is less than 0.1% of the maximum permissible limit of 4 W/kg for wrist-worn devices.
To evaluate the performance of the developed 868 MHz wireless sensor platform, the wireless communication range measurements were conducted in an indoor office environment. The measured results demonstrated a communication range of approximately 31 m for the 868 MHz sensor platform, which is approximately four times greater than the commercial Fitbit Charge 3 BLE wristwatch device.
Furthermore, the implementation of the developed wristwatch sensor platform and the clinical trial results are performed. The design of the gateway, which is used to forward the wristwatch sensor data to a smartphone app, is also reported. The developed wrist-worn sensor platform was effectively implemented for SpO2 and heart rate measurements of the patients in a clinical setting. The successful clinical trials confirm the potential of the developed sensor platform in future wearable health monitoring IoT applications. The future research work will focus on eliminating the use of a smartphone so that the gateway can directly communicate with the cloud.