1. Introduction
An embedded system is a unique, specific, and highly customized computer system. It consists of a microcontroller and a few input/output devices. Inside the microcontroller are the central processing unit (CPU), program and data memory, and input/output peripheral components. Those components provide simple architectures for a program to execute unique and specific tasks. Additionally, an embedded system has the characteristics of high reliability, less power consumption, easier implementation, being lightweight, and lower cost. Simple architectures with many characteristics attract developers to develop diverse applications for consumer electronics, communication and control industry, and smart appliances.
Simple architectures and characteristics, and diverse peripherals and input/output devices, create various embedded systems such as for temperature, humidity, or light control. Frequently used peripherals have a universal asynchronous receiver/transmitter (UART), serial peripheral interface (SPI), interintegrated circuit (I2C), or secure digital-input/-output (SDIO) interface. They can be used to develop infrared-distance photoelectric data collection, collision-event avoidance, light on/off control, or object-tracking devices [
1]. Each embedded system combines one, two, or more peripherals into its design, corresponding to the design specifications. One real-life embedded system is presented in [
2] that comprises a processor, memory, touch screen, display, and UART. It is an embedded e-book system with ARM9 CPU, 32 MB flash memory, a 3.5-inch touch screen, and audio UART. It reduced the design size and cost to smaller and cheaper than those of a personal computer. That is, each component in an embedded system is necessary while tasks are being executed. Consequently, one embedded system does not have any redundant components unless the function of a component is integrated into a microcontroller.
Figure 1 shows diverse embedded systems with various peripherals to connect to miscellaneous sensors. One embedded system is demonstrated in
Figure 1a that consists of a microcontroller with UART and I2C peripherals. The UART peripheral is generally used to implement RS-232 to connect two devices. It limits scalability due to merely connecting two devices. If an embedded system needs to connect more devices, a solution is adopting the I2C peripheral. Another embedded system is presented in
Figure 1b that comprises a microcontroller, and UART and SPI peripherals. The SPI interface is faster than I2C is. Another embedded system is illustrated in
Figure 1c that is composed of a microcontroller, UART, SPI, and SDIO. The SDIO interface can not only connect more devices. It also extends communication by wired or wireless protocols.
Figure 2 shows four applications adopting the aforementioned embedded systems in a smart home, office, or space.
There is a rising number of embedded systems that integrate sensors to work in a smart home. According to sensor market trends, it is expected that a massive number of sensors will be deployed in future households. Therefore, Klemenjak and Elmenreich [
3] aimed to analyze user behavior and energy consumption. They presented an open-hardware energy measurement approach to reflect the power consumption of a certain appliance and impact on the environment. Visutsak and Daoudi [
4] addressed smart home technology for the elderly, and proposed a specific smart home model with a selection of passive and active-intervention devices. In order for devices to communicate with each other, the Internet of Things is usually adopted as the network technology. Florea et al. [
5] addressed several standardized protocols with diverse networking levels on embedded devices to achieve low memory, processing power, and data rate. Chen et al. [
6] also applied the Internet of Things to interconnect and have embedded systems and smart devices to collaborate in cyber–physical systems. In the smart-sensory-furniture (SSF) project of ambient assisted living, Bleda et al. [
7] presented an SSF sensor layer for sensing massively distributed objects with energy limitations and other factors. Researchers such as Chen et al. [
6], and Bleda et al. [
7] expressed a need to process the awareness information from pervasive sensors because they adopted Internet of Things technology. Lalanda et al. [
8] defined a self-aware solution relationship mechanism and proposed context-management software in a service-oriented pervasive environment. As opposed to software, Adiono et al. [
9] presented prototyping for controlling devices in smart homes. Considering the shift from the smart home to an overall smart space, issues relating to interconnectedness, collaboration, monitoring, management, control, or power consumption have become more complex. Zeng et al. [
10] proposed a system-level design approach for smart spaces that constrained cost and power consumption. Regarding it as a multiobjective issue, Deuri and Sathya [
11] proposed the cricket-chirping algorithm, and validated their solution using multiobjective test functions. It was used to solve the disc brake and weld beam design problems. As a problem from multiple object functions evolves to multiple levels, Dutta and Datta [
12] applied a combine-and-transform method to combine both levels of a multiobjective optimization problem to a single level. Zhou [
13] presented a decomposition-based multiobjective tabu search algorithm for multiobjective unconstrained binary quadratic programming problems. The procedures included uniform weight-vector collection decomposed into an aggregation function set and tabu search. Experiment results showed that the proposed solution was effective in meeting its benchmarks.
2. Problem Formulation
A smart office usually splits spaces into business, work, or meeting room areas. A smart home has a few spaces, such as the living room, kitchen, bedroom, and dining room. These spaces generally have some embedded systems set up for collecting data, detecting smoke, or tracking objects for office protection or home care. From an architecture viewpoint, those embedded systems form distributed embedded systems with wireless services in different spaces to the communication center for transferring data. However, interfering substances such as walls, doors, furniture, beam columns, and electromagnetic radiation implicitly or explicitly affect communication strength between embedded systems and the communication center, resulting in each embedded system needing more electromagnetic wave power density for data transfer. Consequently, locations in different spaces of distributed embedded systems and the communication center, called the embedded system layout, require a good arrangement for better electromagnetic signal strength and less electromagnetic wave energy consumption. Moreover, in focusing on an embedded system layout, we assumed the following. First, due to electromagnetic radiation, electromagnetic wave power density through walls is classified into semidirect, indirect, or/and subindirect effects. Second, wall reflection and absorption were ignored because they are hard to identify. Third, electromagnetic wave power density was generated by the communication center and embedded systems. Lastly, the measured electromagnetic wave power density included the transmitter and received data.
4. Experiment Results
Figure 3 illustrates the smart office with business, work, and meeting room areas where a communication center
C and 16 embedded systems S
0, S
1, …, and S
15 were set up. We implemented embedded systems with a wireless function in an Arduino WeMos D1 miniplatform [
16]. The evaluation tools included hardware and software. Hardware was the Tenmars TM-196 instrument [
15], Fortinet FortiAP-221C [
17], and mobile power bank [
18], which were used to measure electromagnetic wave power density, serve wireless service, and supply power for Arduino embedded systems, respectively. We developed MROL application software to detect and assess the communication quality among the communication center and embedded systems.
According to the locations of the 16 kinds of embedded systems in
Figure 3, four candidates of the communication center, namely,
C1,
C2,
C3, and
C4, were used to assess the proposed multiobjective optimization embedded system layout.
Figure 10 exhibits the first test case
C1 at the business area. The measured results of electromagnetic wave power density and signal strength are shown in
Figure 11 and
Table 2, respectively. The Dist. column represents the Euclidean distance from
C1 to the embedded system. The -dBm column represents the communication strength between
C1 to the embedded system. The P.D. column represents electromagnetic wave power density. The first test case was mainly used to assess the
n-shaped layout effect. For S
0, it was set at local effect. Consequently, it had the best communication strength, but consumed the highest electromagnetic wave power density. The measured data for communication strength and electromagnetic wave power density in
Table 2 were 34-dBm and 2500 × 10
−6 W/m
2, respectively. The significant value of 2500 × 10
−6 W/m
2 was because the location was the nearest to
C1. S
1, S
2, S
3, S
6, and S
7 were located in the direct-effect area. Those embedded systems had better communication strength and electromagnetic wave power density. The measured results were between 39 and 49-dbm for the former, and 30 and 55 × 10
−6 W/m
2 for the latter. S
8, S
9–S
12, and S
14 worked in the semidirect-effect area. Communication strength progressively decreased, corresponding to the distance. According to the experiment, communication quality was worse while the communication strength value was greater than 65-dBm. As a result, S
10–S
12 were the candidates to rearrange the locations. S
12 measured electromagnetic wave power density to be 40 × 10
−6 W/m
2. This was different from S
10 or S
11, perhaps because another unknown communication center was located in another building close to S
12. Other embedded systems S
4, S
5, S
13, and S
15 were classified into an
n-shaped layout with a door. In order to distinguish the effects from
n-shaped layout, the effect column in
Table 2 was labeled in lowercase. S
4 worked in the direct-effect area to gain better communication strength than that of S
5. However, it had higher electromagnetic wave power density than that of S
5 due to the direct path of the electromagnetic wave through the door. S
5 was located in the semidirect-effect area with weaker communication strength than that of S
4. S
13 was located in the indirect effect area with a communication strength of 60-dBm. S
15 was located in the subindirect-effect area with weak communication strength because of a value greater than 65-dBm. Considering electromagnetic wave power density, all embedded systems consumed 3116 × 10
−6 W/m
2.
The second test case is demonstrated in
Figure 12, where
C2 was set as the work area. The measured results of electromagnetic wave power density are demonstrated in
Figure 13. This test case was used to evaluate the
n-shaped layout with a door. In order to distinguish the effect from that of the
n-shaped layout, the effect column in
Table 2 was labeled in lowercase. S
4 and S
5 were located in the local effect area, resulting in the best communication strength gained, with 33 and 32-dBm, respectively. S
5 consumed the most electromagnetic wave power density, 7000 × 10
−6 W/m
2, in comparison to other embedded systems due to its location being the nearest to
C2. S
1 was located in the direct-effect area that had better communication strength than that of S
0. In comparison, with electromagnetic wave power density to S
0, S
1, and S
0 consumed 50 and 100 × 10
−6 W/m
2, respectively. S
0, S
2, S
3, S
6, S
7 to S
12, and S
14 were located in the semidirect-effect area with values of communication strength from 57 to 74-dBm. According to the experiment, communication quality was worse while the value of communication strength was greater than 65-dBm. Therefore, S
11, S
12, and S
14 were the candidates to rearrange the locations. S
13 and S
15 were located in the subindirect-effect area with worse communication quality, resulting in them becoming candidates to rearrange the locations. Considering electromagnetic wave power density, all embedded systems consumed 7800 × 10
−6 W/m
2.
Because the first and second test cases had few candidates for rearrangement, the third and fourth test cases aimed to decrease the number of rearrangement candidates to improve the embedded system layout.
Figure 14 shows the third test case that set up
C3 at the S
6 location. Embedded systems except for S
4, S
5, S
13, and S
15 are discussed with regard to the
n-shaped layout. From the viewpoint of -dBm in MROL in
Table 3, values from 33 to 56 indicated that those embedded systems had better communication strength. On the other hand, considering S
4, S
5, S
13, and S
15 with an
n-shaped layout with a door, only S
15 had worse communication quality. Only one embedded system needed rearrangement. Considering electromagnetic wave power density, it consumed 3950 × 10
−6 W/m
2 for all embedded systems; measured results are illustrated in
Figure 15. The final test case set
C4 at the S
3 location. Embedded systems except for S
4, S
5, and S
13 are discussed with regard to the
n-shaped layout. The values of -dBm in MROL ranged from 35 to 73. There were three embedded systems, S
10–S
12, with worse communication strength that needed to be rearranged. Embedded systems with an
n-shaped layout with a door had efficient enough communication strength to work.
Considering electromagnetic wave power density, all embedded systems consumed 3995 × 10
–6 W/m
2; measured results are illustrated in
Figure 16. On the basis of all test case results, the third test case had better electromagnetic wave strength and less of an electromagnetic effect.
5. Conclusions
Embedded systems with smaller, cheaper, and easier-to-implement characteristics are more popular for designing smart objects, devices, applications, or services. Recently, there has been an increasing number of functions, such as wired, wireless, or/and sensor peripherals that are integrated into embedded systems to provide interconnected and collaborating services. As diverse embedded systems are continually set and joined up to smart offices, either electromagnetic wave power density or the communication strength is a significant issue considering energy-saving or cooperation work for all embedded systems. This work studied the layout of multiple embedded systems for communication strength and electromagnetic wave power density optimization. Our prior works measured the original data of electromagnetic signal strength and power density for smart offices, shown in
Figure 3. For the issue of electromagnetic signal strength, we analyzed deployment topology into an
n-shaped layout and an
n-shaped layout with a door in smart offices. In the
n-shaped layout, we classified electromagnetic signal strength into local, direct, and semidirect effects, and their scope. Then, we defined a set of formulas to determine the effect of each embedded system. We presented local, direct, semidirect, indirect, and subindirect effects for an
n-shaped layout with a door. Each effect was derived from the scope, and then defined a set of formulas. Those formulas can help users to quickly identify the scope of each effect while the communication center is set up. In the experiments, we measured each effect in either the
n-shaped layout or the
n-shaped layout with a door for four test cases. Experiment results indicated that the local effect had the best communication strength, but electromagnetic wave power density was critical. The direct effect had better communication strength and less consumption of electromagnetic wave power density than the semidirect effect did. By comparing the semidirect and indirect effects, the former generally had better communication strength than that of the latter. Regarding the subindirect effect, it consumed the least electromagnetic wave power density, but communication strength was the weakest among all effects. Lastly, a summary of the experiment results pointed to diverse effects with the optimization layout for embedded systems.