1. Introduction
Nearly 71% of the Earth’s surface is covered by the ocean, a connected body of water that is usually split into different principal oceans and miniature seas. Ocean warmth influences environments and weather patterns that change life on earth. Freshwater in lakes and rivers represents less than 1% of global surface area. The health of the ocean determines the health of the planet. The Internet of Underwater Things (IoUT) is described as a global network of intelligent, interconnected underwater things that allows controlling large unexplored water areas [
1]. These things can be underwater sensors, autonomous surface vehicles (ASVs), ships, autonomous underwater vehicles (AUVs), etc. The IoUT can be very useful in many practical applications such as environmental monitoring, marine research and disaster prevention. Also, it could be a key technology in implementing a future smart underwater city. The IoUT was discussed in [
2], and its importance was emphasized in previous work [
2,
3,
4].
Mainly, four communication topologies can be used in IoUT [
5,
6]: optical, electromagnetic, magnetic induction and acoustic wave. Due to natural spreading problems such as scattering and line of sight (LOS), optical underwater communication its applications are restricted to be used only in clean water. Harsh underwater conductivity attenuates the electromagnetic and magnetic induction waves and restricts its operating frequency to be at a very low-frequency band. Acoustic is the widest topology used in underwater applications, where it achieves a wide coverage range which can be on the scale of kilometers, but unfortunately it is a high-power communication system with a limited channel bandwidth. The difference among these different physical underwater technologies in terms of features, pros and cons has been summarized in [
6]. Power consumption represents the main IoUT challenge, due to the recharging capability problem of sensor nodes deployed in the underwater areas which suffer from the lack of a sustainable power source to feed their batteries. Many researchers have tried hard to increase energy and spectral efficiency in underwater/underground communication based on an effective magnetic induction physical layer technology [
7,
8,
9,
10,
11]. They suggested magnetic induction communication as a physical layer for the internet of underwater/underground things [
7,
8,
9,
10,
11] due to its shorter propagation delay and highly environment-independent channel behavior. However, unfortunately it still has many challenges to be applicable for the IoUT due to the short communication ranges of magnetic induction (10–100 m) compared to the km scale of acoustic communication. Also, its effective bandwidth and channel capacity are practically quite limited in most underwater transmission scenarios, within 100 KHz and 100 Kbps, respectively [
7]. The reasons behind that are related to the sensitivity of a coil’s optimal operating frequency where small deviations can cause power reflections leading to a narrower effective bandwidth, and also to the operational limitation in low frequency in order to mitigate the high attenuation caused by eddy currents in the highly conductive water [
11]. Hence, acoustic communication is still the most usable of the physical layer technologies which could be used in the IoUT. However, unfortunately according to the authors’ knowledge, there is no research which has addressed the IoUT energy saving problem in acoustic physical layer technologies regarding modulation techniques.
Multiple-input multiple-output (MIMO) technology represents the underwater communication key as it offers an increase in capacity and diversity gain for acoustic underwater wireless networks [
12]. Unfortunately, MIMO has many limitations, such as implementation complexity with a high number of antennas, and its performance deterioration in multipath propagation environments such as the underwater channel. So, in many MIMO systems power loss problems lead to the implementation of a tiny number of transmitter antennas [
13]. Hence, the spatial modulation (SM) scheme has been proposed where just a single transmitter antenna is activated in each communication, and a spatial domain is used to compensate the possible data rate [
14].
SM is considered as one of the space modulation techniques (SMT), where one or more antennas can be selected for transmitting information. It provides a higher data rate than the traditional single-input-multiple-output (SIMO) systems, reduces its complexity without antenna synchronization, while avoiding inter-channel interference (ICI). Different SMTs have been introduced [
13,
15,
16,
17,
18,
19,
20,
21,
22,
23], extending the SMTs principles from enabling just one radio frequency (RF) chain at the transmitter side into enabling one or more RF chains with different techniques in order to achieve high throughput, enhanced signal-to-noise ratio (SNR), or both. One of these powerful techniques is generalized spatial modulation (GSM) [
21,
22]. In GSM, multiple antennas are active in each communication in order to improve the achievable data rate. Those activated antennas transmit the same symbols using the same data constellation for all symbols. Consequently, the incoming data is conveyed via the indices of the activated transmitter antennas and the modulated information symbol. Therefore, the available data rate of GSM is higher than that of the current SM because of the activation of a greater number of transmitter antennas. Based on the GSM mechanism, fully generalized spectral efficiency (FGSM) was proposed in [
15,
16], and the antenna subsets vary from only a single activated antenna, to multiple/all activated antennas. The variation of the number of transmitter antennas enhances the realization of the communication channel and therefore increases the available data rate.
As the data signal constellations can be used as an additional dimension to map the information bits [
17], the transmitted data can be carried not just by the indices of the transmitter antennas as in SM, GSM, and FGSM but also by using the signal constellation type. The primary and secondary constellations are used to increase the number of combinations between modulations and transmitter antennas. A significant performance gain of spectral efficiency, receiver complexity reduction, and power saving are achieved in this way compared to conventional SMTs. Based on that, in this paper, a new scheme called enhanced fully generalized spatial modulation (EFGSM) is proposed. In this new scheme, the data signal constellation is used as an index, and information is conveyed through a varied number of transmitter antennas. At least two transmitter antennas are activated to transmit information. One or more are based on the primary signal constellation and other one or more are activated based on the secondary constellation. The numerical and simulation results show that a significant performance gain in terms of energy and spectral efficiency, receiver complexity and average bit error rate (ABER) is achieved using the proposed scheme compared to conventional FGSM, GSM, and SM. Our main contributions can be summarized as follows:
- (1).
We investigate for the first time the power consumption of the IoUT, and a new energy efficient communication system for the IoUT is proposed based on a modification of the fully generalized spatial modulation. The key idea of the proposed scheme uses the data signal constellations as an additional dimension to map the information bits by combining the enhanced SM and FGSM.
- (2).
Using the well-known union boundary technique, we determine closed-form ABER approximations for the proposed EFGSM in the acoustic IoUT. Furthermore, inclusive Monte Carlo simulations are applied to validate the derived formula.
- (3).
We calculate the power saved using the proposed EFGSM scheme in reference to the quadrature amplitude modulation (M-QAM) and other SMTs which can be used as a modulation scheme in the acoustic IoUT.
- (4).
We analyze and evaluate the receiver’s computational complexity of the suggested EFGSM by defining the total number of real operations (TNRO) needed at their maximum likelihood (ML) decoders. Furthermore, it is compared with the computational complexity of the other current SMTs.
The rest of the paper is structured as follows:
Section 2 represents the relevant literature. In
Section 3, we explain the proposed EFGSM system model. The performance analysis of the EFGSM is introduced in
Section 4.
Section 5 presents the simulation results, and finally the paper is concluded in
Section 6.
2. Related Work
In recent years, spatial modulation techniques (SMTs) such as SM [
18,
19], GSM [
21,
22], quadrature spatial modulation (QSM) [
23], FGSM [
15,
16], and fully quadrature spatial modulation (FQSM) [
16] have been considered as promising MIMO techniques that achieve better performances in terms of spectral and energy efficiency as well as in reducing system complexity in comparison with the conventional MIMO counterparts.
In SM [
18,
19] one single RF chain, out of the
antennas, is activated to convey the data constellation symbol, and the index of this active RF chain carries additional data bits. The available data rate on the traditional SM technique
is relative to signal constellation
-ary and the number of transmitted antennas
. It allows poor communication since just one transmitter antenna is working at any given time. Consequently the SM achievable data rate,
, can be expressed as follows [
18,
19]:
where
M represents the modulation order of the conveyed symbol.
The GSM [
21,
22] was proposed to improve the SM achievable data rate by activating multiple antennas. In GSM, the same data is transmitted through two or more antennas using the same data signal constellation. However, the data rate improvement is marginal, moreover, activating more than one antenna without efficient use will rapidly consume the energy source. The GSM achievable data rate,
, is considered as proportional to the logarithm of the binomial coefficients, and can be written as follows [
21,
22]:
where
indicate the floor operator, the binomial coefficient, and the number of transmitting active antennas, respectively.
The spatial constellation diagram was extended into two orthogonal dimensions in QSM where real and imaginary parts are transmitted through one or more transmitter antenna [
23]. The real part of the data constellation symbol is transmitted via one dimension, and the imaginary part is transmitted via another dimension. By this data bit transmission method, the achievable data rate is enhanced compared to SM and GSM, but regrettably, the enhanced data rate is logarithmic with the square of the number of
. Accordingly, the QSM achievable data rate,
can be expressed as follows [
23]:
In enhanced SM (ESM) [
13,
17], the information bits are not only transmitted by the active transmitter antennas index(es) but also by using the signal constellation type. Hence, an integer of the power of two transmitter antenna
out of the transmitting antennas can be selected to be used in information transmission. Half of the activated transmitter antennas are forced to transmit information based on the primary signal constellation, and the others transmit information using the secondary signal constellation which is obtained by a single-step geometric insertion among the points of the primary signal constellation. The points of the secondary constellation are located at the midpoints of the intersections created by neighbor points of the
-QAM which are used as a primary constellation [
17]. For example, if there are four transmitter antennas used, the maximum number of potential working antennas is two to transmit antenna conveyed information, one activated by the primary signal constellation, and the other is activated by secondary signal constellation. The ESM marginally increases the achievable data rate with significant improvement in the average energy per transmitted symbol
.
In FGSM, the number of active transmitter antennas is used as an index to transmit data by the same data signal constellation [
15,
16] where the active antennas indexes are used to send information bits, and also their varying quantity r is used as an index to carry additional information bits. For example, one or more antennas can be activated to send data using the same data signal constellation, hence the number of activated antennas carries information as well. In other words, it is a combination of SM and GSM where if one antenna is activated, FGSM is similar to SM, while if multiple/all transmit antennas are activated, the FGSM is equivalent to GSM. In this way, FGSM achieves greater data bit transmission, and its achievable data rate
is given by:
As in the FGSM combination technique, the varied number of active antennas has been applied to QSM in [
16] resulting in FQSM. FQSM transmits the real and the imaginary components of the signal constellation in one or more transmitter antennas in orthogonal dimensions, and its achievable data rate
is expressed as follows:
Despite the achievable data rate enhancement in FGSM and FQSM compared with conventional schemes, transmitting the same data by multiple transmitter antennas increases the average energy per transmitted symbol which effects power saving dramatically.
3. Proposed EFGSM
The block diagram of the proposed EFGSM including the modulator and demodulator is shown in
Figure 1. The incoming data bits are partitioned into two groups. The first group contains
bits that are referred to as data bits, where
and
stand for primary and secondary signal constellations, respectively. This group is used for modulating the signal constellation symbol from two signal constellation diagrams as designed in ESM-Type1 in [
17]. The second combination of bits represents the spatial bits; it is applied to choose the antenna subset utilized in transmitting information of the constellation symbol. The antenna subset of EFGSM differs from the case of activating just one transmitter antenna to the case of activating
transmitter antennas for each the primary and secondary signal constellations. This contradicts the approaches of the current FGSM introduced in [
15,
16] and ESM-Type1 in [
17], in which a varied transmitter antenna number is activated in the case of FGSM, and different signal constellations are used in case of ESM-Type1. The variation of signal constellations and transmitter antennas enhance the underwater acoustic channel utilization, reduces the average energy per transmitted symbol, and obviates the BER deterioration. The proposed EFGSM can be considered as a combination of ESM and FGSM. Therefore, the achievable data rate of the proposed EFGSM can be expressed as follows:
where
represents the number of active antennas for both primary and secondary constellations.
Different subspaces based on different number of antennas utilize different combinations of active antennas. In all of these subspaces one/multiple active antennas transmit symbols based on the primary signal constellation, , while the other active/multiple antennas transmit symbols based on the secondary signal constellation, . The group of the spatial bits , is used to select the combination of active antennas. Generally, we can say that the transmitted information bits are based on three groups: the primary signal constellation, the secondary signal constellation, and the active antenna indexes. First, the information bits are carried via the primary constellation with an achievable bit rate of . Secondly, the information bits are conveyed using the secondary signal constellation with an achievable bit rate of . Finally, the utilization of the spatial combination of antennas and modulation results in an achievable bit rate of . Unlike the conventional FGSM, using primary and secondary signal constellations increases the number of antenna/constellation combinations. Therefore, EFGSM increases the achievable data rate and decreases energy consumption without altering system complexity. For example, in case of using 4-QAM as a primary signal constellation and binary phase shift keying (BPSK) as a secondary signal constellation, the gutted achievable data rate is 7 bit per channel used (bpcu), on the other hand, the achievable data rate of the conventional FGSM using the same MIMO configuration is only 5 bpcu. However, when using 8-QAM as a primary signal constellation and QPSK as a secondary signal constellation, with the same number of transmitter antennas, the achievable data rate will increase to 9 bpcu in the proposed EFGSM and only 6 bpcu in case of the conventional FGSM.
For a better explanation of the proposed EFGSM transmission procedure, assume, without loss of generality,
=
= 4 as shown in
Figure 1, where
represents the receiver antennas. Using 4-QAM modulator as a primary signal constellation type, the secondary signal constellation is equivalent to the throughput of half of the primary signal constellation, such as BPSK. In this case, the antenna indexing of the proposed EFGSM can be represented as shown in
Table 1, for 7 bpcu. For illustration, in the example mentioned, consider the following generated bits,
, are needed to be transmitted. The first three bits [010] represents the bits carried via the primary and secondary signal constellations. [01] is equivalent to
and conveyed using the primary signal constellation, and the other part [0] (equal to the half of primary signal constellation bits), equivalent to
, transmitted via the secondary signal constellation. The remaining four bits, [1010], are transmitted via the antenna index
. The secondary signal constellation bits will be transmitted over the active antenna,
, and primary signal constellation will be transmitted over the active transmitting antenna
. Therefore, the
transmitted vector of the EFGSM is
. That vector of the EFGSM is sent via
x
over an uncorrelated underwater acoustic channel, H contaminated by an additive white Gaussian noise with
.
The received signal
at receiver side is given by:
and,
where,
and
indicate the summation of columns of activated antennas channel required to transmit the primary signal constellation and secondary constellation, respectively.
,
= 1, 2 …
with
denotes the ceiling operator and
denotes the
columns of channel H.
The knowledge of the channel at the receiver side is assumed to be perfect. Therefore, the maximum-likelihood (ML) decoder can be used in the EFGSM system and is expressed as:
In the ML decoder, the indexed information bits are recovered using the estimated antenna indices combination and the primary and secondary estimated data symbol constellations.