A Universal Multimode (Acoustic, Magnetic Induction, Optical, RF) Software Defined Modem Architecture for Underwater Communication

In this paper, a Universal Underwater Software Defined Modem (UniSDM) architecture is proposed that may operate in different modes (acoustic, magnetic induction, optical and RF), in order to utilize the advantages of each mode and accordingly satisfy the requirements of many latest use cases in underwater communication systems. A detailed description of the novel UniSDM architecture is presented first. The novelty of this architecture is its flexibility, i.e., allowing the designers to produce a device that may include any type of modes operating seamlessly and jointly by exchanging data, control and synchronization. Many challenges, including high system costs and coordination between different modes, are addressed in the paper. Moreover, numerical evaluation is conducted to assess the performance of the proposed UniSDM architecture. Finally, the performance evaluation shows that the utilization of the UniSDM allows to decrease the transmission latency and improve the energy efficiency, while maintaining high reliability and robustness in underwater communication systems.

more.Tethered solutions would be expensive/inflexible and would require very advanced technologies for deployments especially in deep waters.Most importantly, even in shallow water and areas already equipped with infrastructures, like underwater Oil & Gas facilities, tethered solutions are impractical.Thus untethered solutions are preferable in terms of simple usage, higher flexibility, and lower risk for possible damages or intervened by some existing structures.Similar observations apply to complex submerged structures like shipwrecks or coral reefs.These advantages are still very relevant, even if they come at a cost of the energy problems as the power must be supplied by underwater chargers.
Acoustic communication is the most popular physical layer mode due to its reliable and high transmission range communication in underwater systems, which can cover several kilometers.However, acoustic communication has high propagation delays due to the low speed of sound, ≈1.5 mm/µs in water.Furthermore, long-range acoustic propagation is efficient over a limited spectral bandwidth [2], [3], [4], limiting the channel capacity even when advanced technologies such as cooperative MIMO are used [5], [6].High bandwidth acoustic communication can achieve data rates of 1 Mb/s for distances up to 100 m.This makes acoustic communication a preferred mechanism in underwater applications compared to magnetic induction (MI) and optical communication in terms of capacity, however, the propagation delay in acoustic communication is still limited by the speed of sound.
Alternatively, optical communication has the highest propagation speed of 225 m/µs and ultimately wide bandwidth (hundreds of THz).However, optical waves suffer from high absorption and scattering depending on the water characteristics, which limit the achievable transmission range to ≈500 m in the most transparent ocean waters [7], and to ≈10 m and below in turbulent waters.Practical implementations of optical communication systems have achieved up to 10 Mb/s at 150 m using wide-beam LED modems [8], [9].Higher data rates such as 500 Mb/s at 100 m [10] and 5.5 Gbps at 26 m [11] are achieved using a collimated laser beam in a laboratory environment.
Radio frequency (RF) communication is another transmission mode that can maintain a low-range signal propagation and can provide high data rates.Since EM waves in water are absorbed at a distance of the same order of the wavelength, they are either used at close contact, or at very low frequencies, which lead to giant antenna sizes and very low data rates.
Finally, MI allows reliable short-range communications depending on water conductivity with high propagation speed at 33 m/µs [12], [13].It is well-understood that the features of different communication modes are complementary.In other words, one transmission mode can be preferred over the others depending on the transmission requirements, water characteristics, and transmission range.
The contributions of this paper are: • Based on existing design principles of the software defined radios (SDRs), we propose a novel universal software defined modem (UniSDM) architecture [1] that integrates acoustic, MI, optical and RF communication paradigms to realize multi-mode device operations.The UniSDM architecture design can significantly improve the system performance by allowing simultaneous and joint operations between different communication modes, improves synchronization accuracy, and reduces device cost and size.• We introduce algorithms for simultaneous operations of these heterogeneous (acoustic, MI, optical and RF) modes.We also propose performance metrics that can be used for the selection of the appropriate modes based on the use cases and underwater channel conditions.• We present a numerical performance evaluation of an underwater network with UniSDM devices and demonstrate its gains compared to single-mode device networks.The rest of the paper is organized as follows.In Sec.II we review the state of the art related to the UniSDM.In Sec.III we describe the architecture of the UniSDM.In subsection III-B we discuss the flexibility of the UniSDM operation due to its SDR nature.In subsection III-C we explain the criteria and propose an algorithm to be used for mode selection in the UniSDM.Sec.IV consists of two parts.The first subsection IV-A explains the features of the acoustic, MI, optical and RF modes to allow the reader to understand when one mode should be preferred over another.Then subsection IV-B provides the evaluation of the gain that the multi-mode approach can give to the communication quality metrics when applied to some specific network topologies.Finally, Sec.V summarizes the main results and gives the conclusions.

II. STATE OF THE ART
In the following, we discuss the state of the art related to multimode underwater communication, in order to position the novelty of our contributions.Indeed, the integrated multimode systems have been studied in the literature in recent years [14], [15], [16], [17], [18].However, existing solutions mainly use separate modes, e.g.acoustic communication mode for low-bandwidth control signaling, and optical/MI mode for high-speed data communication in short distances.
In particular, in [17] and [18] an underwater fixed sensor network is proposed, which uses the acoustic mode for signaling, while autonomous underwater or surface vehicles use the optical mode for collecting data from the distributed fixed sensors.
Multimode underwater communication were considered in the context of routing algorithms' studies [19], [20], [21].The MURAO routing protocol [21] is proposed for acoustic and optical modes, where the protocol measures the propagation delay in terms of the number of hops, disregarding the propagation speed difference of ≈150000 times between acoustic and optical modes.Accordingly, acoustic paths are often selected by this protocol, although including the information of the different propagation speeds would easily result in multihop paths preferring optical mode, due to its higher propagation speed.A multi-mode underwater network slicing solution was proposed in [22], but no discussion on the modems' hardware was provided.
To the best of our knowledge, so far in the literature little attention has been given to distinctive features of the modes, concerning the maximum range, bandwidth and propagation delay simultaneously.
Moreover, the use of the separate modems has some drawbacks: • Use of separate modems for each mode adds delays when switching between modes.• Further delays may be introduced when packets in each node are forwarded from one mode to another.• The processing is limited to packet reception and transmission.This prevents the implementation of some technologies, e.g.amplify-and-forward techniques where physical layer synchronization and digital signal processing between the modes are needed that will allow to pass the non-processed signals from one mode to another.• Inter-mode cooperative operation is difficult, since it requires precise synchronization (e.g.symbol and carrier phase synchronizations).Note that an integrated acoustic and MI software defined modem architecture was recently studied in [23].They present a proof-of-concept of a cooperative MIMO system whose nodes are implemented by using off-the-shelf SDRs.The paper has the following shortcomings: 1) Only two modes (acoustic and MI) are considered.
2) The implementation does not consider the reconfigurability, flexibility or the possibility to provide additional modes.3) They only implement cooperative 2 × 1 MISO acoustic data transmission supported by synchronization of time and frequency in MI mode.

III. THE UNIVERSAL SOFTWARE DEFINED MODEM ARCHITECTURE
We propose the UniSDM architecture that enables joint/simultaneous operations of multiple modes, including cooperative communications, utilizing two or more modes at the same time.The proposed UniSDM has a universal physical layer (PHY) that combines functions of the separate modes, and provides synchronization and signal processing.Hence, it is capable of supporting any physical front-end or several front-ends simultaneously.
The advantages of the UniSDM are:

A. The Architecture
Before presenting the detailed description of our UniSDM architecture, we underline a number of features that it supports.In fact, our architecture aims at solving a number of tasks that arise in modern communications: • The UniSDM supports multimode operation, thus it has several physical modes front-ends with their specifications.• The UniSDM is flexible and supports various existing and future communication systems.Because of that, it is based on SDR that uses digital signal processing (DSP) for signal synthesis and reception, including implementation of media access control (MAC) protocols, error correction (e.g., forward error correction (FEC), frame check sum computation, etc.), and modulation schemes.• The UniSDM modules may be distributed: some applications may require the use of centralized multi-antenna systems.An example of such a system is a shore communication gateway that provides connection to a network along the coastal line.It may have many mode frontends (MFs) allowing multiuser MIMO operations in acoustic bands and handover between antennas for short-range communications (optical and MI).
We show different types of MIMO operations which can be realized by the UniSDM architecture.Unlike traditional MIMO devices that use multi-channels within the same frequency band and same medium (terrestrial, air, underwater, soil, etc.), in the UniSDM design, the modulation and coding schemes (MCS) may combine different frequency bands.Thus, our proposed MIMO operating mode within the UniSDM design can be either homogeneous or heterogeneous, as well as cooperative or not cooperative.In the following we explain the details.
• The classical MIMO mode is homogeneous and noncooperative."Homogeneous" means that the MIMO may use only a single mode (RF or optical or MI or acoustic) and the same frequency band for all communications.• The key idea of the cooperative MIMO (Co-MIMO) is using a cluster of several independent devices as an antenna array to improve the communication rate, range and energy efficiency.To transmit a packet in a coordinated manner, the devices must first distribute the data to be transmitted among themselves and accordingly synchronize their clocks [5], [6].
An example of the Co-MIMO application would be a communication between a swarm of UAVs (unmanned underwater vehicles) and a remote control center.Each device may have only one transceiver, but they jointly work as a massive antenna array.The Co-MIMO achieves high beamforming gain and improves energy efficiency while maintaining low hardware device complexity.Homogeneous cooperative MIMO means that only one mode (acoustic or RF or MI or optical) and the same frequency band are used for both intra-and inter-cluster communications.
Although homogeneous Co-MIMO improves the performance metrics, it has considerable communication overhead to establish synchronization between the devices.This will impair the actual data communication.For example, during the synchronization within the cluster, no data can be transmitted or received from/to other clusters.Because of the low spreading factor (typically 1 to 2) of the acoustic propagation, the intra-cluster communication can also create interference in the long ranges.• To mitigate the drawbacks of the homogeneous Co-MIMO, heterogeneous Co-MIMO can be implemented.
In heterogeneous Co-MIMO, different modes (acoustic, RF, MI, optical) are used for intra-and inter-cluster communications.
Note that for inter-cluster communication we suggest to use acoustic mode due to the long range capabilities, and any other mode (e.g.optical, MI or even a higher frequency acoustic mode) for intra-cluster communication.Both optical and MI modes have relatively short ranges (in the order of 1 m to 100 m) but relatively high carrier frequencies and bandwidths, which may allow fast and efficient communication of the control data.
Consequently acoustic mode can be used for inter-cluster communications.
An overview of the architecture is shown in Fig. 1.The UniSDM is partitioned into two parts: Baseband Unit (BBU) and Multi-Mode Heads (MMHs).These parts may operate as a single unit or can be distributed and connected, similar to modern terrestrial mobile networks for building Cloud Radio Access Network (Cloud-RAN).The connection interface shall provide functionality similar to the Common Public Radio Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
Interface, but account for the network modality, i.e., provide access to various modes with different properties.
A UniSDM instance is comprised of a Baseband Unit (BBU), and one or more Multi-Mode Heads (MMH).The details of the architecture of the BBU and MMH are described in the rest of this subsection.
1) The Baseband Unit (BBU) Architecture: The BBU is the brain of the UniSDM.It is comprised of a BBU controller (BBU-C) and a BBU digital signal processor (BBU-DSP).The BBU-C is responsible for interfacing to user, and controlling all the UniSDM subsystems.The BBU-DSP, on the other hand, implements all the algorithms of the UniSDM, i.e. the protocols from high-MAC to high-PHY.The data flow from the BBU-C through the BBU-DSP to the MMH, for transmission chain, and vice versa for reception chain.
2) The Multi-Mode Head (MMH) Architecture: The MMH provides the BBU with simultaneous access to the available wireless modes.It is comprised of a Signals Unit (SU) and a number of Mode Frontends (MF), i.e., RF, acoustics, MI, and optical MFs.
The SU is the spinal cord of the UniSDM.It provides a number of functions, as follows: • Digital signal processing of the Low-PHY layer.The tasks that require minimal latency or high throughput at low or medium complexity may be offloaded from the BBU to a simple-structured processing unit in the SU.Examples of such processing are signal filtering and frequency conversion.• Analog-to-digital and digital-to-analog conversion (ADC and DAC).Digital electric signal is converted to analog signal before being transmitted through selected frontend(s).Conversely, received analog signal is converted to digital signal before being processed by the SU and BBU. ) are reported to the SU and to the BBU, so that BBU can make decisions on the technologies that can be supported using this MF.
-Magnetic Induction front-ends essentially are lowfrequency RF front-ends with magnetic antennas.Therefore, they share most of the properties with RF-MFs.The significant difference is the antenna type.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.Implementations may include 3-axis antennas to guarantee omni-directional transmission and reception [24].MIMO operations can be implemented with the UniSDM architecture in various ways: 1) Mode Front-ends (MFs) may include multiple channels for receiving and transmitting chains.For example, MF can provide 4 acoustic Rx/Tx channels.
2) The UniSDM may include several identical MFs.For example, the UniSDM may include 4 identical 1-channel optical MFs.The structure of the UniSDM allows coop-eration and synchronization between multiple modes, thus it is possible to use sevaral MFs together to provide MIMO operation.3) These two approaches can be used together, for example, the UniSDM may include 2 MFs, of a specific mode, each with 4 channels, providing 8-channel operations.

B. Operation
A UniSDM instance would typically support multiple operation modes with one or more communication technologies per mode.As an example, RF mode could implement 5G technology for broadband data connectivity or LoRa technology for low-rate data transmission.In the same way, an acoustic mode can implement various communication protocol, e.g., JANUS or proprietary protocols.Depending on required communication QoS, and the measured channel state information for all supported modes and bands within modes, one or multiple nodes are selected to conduct transmission.Due to softwaredefined-modem nature of the UniSDM, the configuration can be updated remotely.This allows supporting future technologies, including the cooperative multimode MIMO.
Since each technology operation requires resources (such as memory for storage, DSP resource for implementation, power for active operation), the mode selection algorithms should take the resources' availability into account.
The technologies may be presented in an UniSDM instance in various states:

C. The Mode Cooperation
To establish communications between two nodes, the UniSDM device shall decide which mode or modes it is going to use.To do that, it does channel probings by different modes, and accordingly establishes a link.Once one or more links are established, the links' channel state information (CSI) is collected at the BBU-C to select one of the modes to perform data transmission based on a given criteria (discussed below).
In short, the UniSDM operation is realized as follows: 1) Probe the available modes and establish communication links.2) Request the transmission, and select a particular mode according to a particular criteria given below.3) Send the data (payload).

1) Probing the Modes:
To probe the modes, the node sends probing packets at maximal available power.When another node receives such packet, it replies to it.During this twoway communication, the nodes measure the received power and response delay, and thus obtain information about the propagation delay and the path loss.
The modes' probing is done sequentially, starting from the mode that provides the lowest delay.This approach reduces the required time, and minimizes interference.For example, if the two nodes can establish optical mode link, they can use it for all types of communication, both control and data.In case the optical link is not available, it should probe a magnetic induction mode.If neither optical nor MI probing succeeded, it should switch to acoustic, that provides the highest range, but at the lowest speed.
2) Mode Selection Criteria: Although switching from one mode to another according to their availability was suggested in [25], we propose aggressive switching to a mode according to one of the appropriate metrics, e.g.: • SNR threshold with higher-speed mode priority.
• Energy efficiency per distance [J/bit/m] -for the network energy savings.• Transmit time -for latency minimization and throughput maximization considering non-uniform node density.
Some other criteria can also be considered, such as: • Energy efficiency of current device [J/bit] -it can be used for a particular node's energy saving, especially when this node is less capable (e.g. has less energy) than the others.• The product of the transmission time and the mode coverage area (or volume) -to maximize per-area or pervolume throughput, • The product of the transmission time and the number of neighbors in the node vicinity -to maximize the per-area throughput taking into account the nodes' distribution in space.
Traditional engineering approach uses separate modes for separate tasks, i.e. acoustic modem for remote control at any distance, and optical modem for data offload at short ranges.We suggest changing this paradigm for several reasons: • Today we can assume that if we are using acoustic modem, our acoustic communication system is the only one in the area, and there is no interference.• When UW sensors, AUVs, ROVs and other UW IoT devices become more widespread, the medium will become crowded.• Automatic switching to short-range technology would mitigate interference problems.

IV. PERFORMANCE EVALUATION
In this section we provide a numerical comparison and analysis for the communication modes' performance.Then, we show the performance of the multi-mode UniSDM for multiple networking scenarios.

A. Modes Comparison
In this subsection, the performance of each communication mode is analyzed.We want to emphasize some important features of the modes that affect the decision of preferring one mode to another.This section is intended to familiarize the reader with the trade-offs between the different modes on the scale of single hops, and to help understanding of the numerical results obtained for the networks.
1) Acoustic: Underwater acoustic communication is the most popular physical layer technology for its reliable and high transmission range communication [2], which can exceed several kilometers.However, it has high propagation delays due to the low speed of sound that is ≈ 1.5 mm/µs in water.Furthermore, long-range acoustic propagation is efficient over a limited spectral bandwidth, limiting the channel capacity even when advanced technologies such as cooperative MIMO are used [5], [6].High bandwidth acoustic is capable of achieving data rates of 1 Mb/s for distances of up to 100 m.It allows it to compete with some implementations of MI and optical channels in terms of channel capacity, but the propagation delay is still limited by the speed of sound.
The acoustic mode model is an approximate model of modem that operates in the frequency band around 30 kHz.The spreading factor is k ac = 3/2 (in between spherical and cylindrical transmission), and the received signal power is computed as: where r is the distance, P TX = 10 W is the transmission power, and S RX = 1 cm 2 is the receiver area.The link is considered for communication only if the received power is higher than a certain threshold: P RX ,ac > P thr ,ac .In practice, this threshold depends on the receiver sensitivity and on the background noise.In the numerical evaluation, this threshold is taken to be P thr ,ac = 10 −9 W which gives the range ≈ 1.5 km, which models the expected communication range of the commercial off-the-shelf acoustic modems.Furthermore, we fix the speed of sound in water to be c ac = 1500 m/s, and the data rate to be b = 50 kbps.
2) Optical: Optical communication has the highest propagation speed of 225 m/µs and ultimately wide bandwidth (hundreds of THz).However, optical waves suffer from high absorption and scattering depending on the water characteristics, which limit achievable transmission range to ∼ 500 m in clearest waters [7], and to ∼ 10 m and below in turbulent waters [26], [27].Practical implementations of optical communication systems have achieved up to 10 Mb/s at 150 m using wide-beam LED modems [8], [9].Higher data rates such as 500 Mb/s at 100 m [10] and 5.5 Gb/s at 26 m [11] are achieved using collimated laser beam in lab environment.

RF communication is another transmission mode that can
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
maintain a low-range signal propagation and provide moderate data rates.
The optical mode model is a model of LED-based underwater optical modem.The spreading factor is k opt = 2, the received signal power is computed as: where r is the distance, P TX = 10 W is the transmission power, S RX = 1 cm 2 is the photodetector area.The attenuation coefficient, which combines the absorption and scattering coefficients c(λ) = a(λ) + b(λ), is dependent on the water type, as follows: • c(λ) = 0.011 m −1 for clearest ocean or pure water at λ = 405 nm [7], • c(λ) = 0.151 m −1 for clear ocean [28], • c(λ) = 0.399 m −1 for coastal ocean [28], and • c(λ) = 2.195 m −1 for turbid harbour [28].The link is considered possible if the received power is higher than a certain threshold: P RX ,opt > P thr ,opt .In the numerical evaluation, this threshold equals P thr ,opt = 10 −10 W which matches the performance of the commercial off-the-shelf modems with no external light interference.Such modems reach up to 150 m in pure water for the models using photomultiplier tubes for detecting the light [8], and around 50 m for models with photodiode receivers [9].Under these assumptions, the model operating range in clear ocean is limited to ∼ 50 m.
The optical channel data rate is fixed to be b = 10 Mb/s, and the signal propagates at the speed of light in water, c opt = 225 m/µs.
3) Magnetic Induction: MI allows reliable short-range communications depending on water conductivity with high propagation speed at 33 m/µs [12].The received signal power is computed as: where r is the distance, P TX = 10 W is the transmission power, d RX ,MI = 1 cm is the characteristic size of the receiver antenna, and α = 1/δ = √ πf c µ 0 σ in the inverse of the skin depth [12].To simplify the model, we neglected the losses due to water conductivity, so α = 0 in the model.
The link is considered possible if the received power is higher than a certain threshold: P RX ,MI > P thr ,MI .This threshold is defined by thermal noise (Johnson-Nyquist noise) at the receiver input, which can be computed as where k B = 1.38 • 10 −23 J/K is the Boltzmann constant, T ≈ 300 K is the receiver temperature, and ∆f = 1 MHz is the frequency bandwidth, thus the noise power threshold are P thr ,MI ≈ 3 • 10 −14 W.Under these conditions, maximal operating range is limited to ≈ 8 m.The MI data rate is assumed to be 1 Mb/s.4) Radio Frequency: Since EM waves in water are absorbed at a distance of the order of magnitude of the wavelength, the RF underwater applications can be classified into two main approaches:  • Very short range range communication, essentially putting the devices into contact.• Use of very low frequencies, which leads to giant antenna sizes and very low data rates.
Hence, the RF based communication is considered impractical for underwater communication.That being said, RF plays a key role in the surface communication of ships, buoys and other vehicles, using a plethora of communication technologies: VHF marine radios, HF radios, satellite communications, mobile communications (from GSM (2G) to 5G) and others.
Since we focused on the evaluation of underwater communications, we did not implement any model of RF communication in our evaluation.
5) Analysis: Figure 4 shows the dependence of the required transmission power versus distance for acoustics, MI and optical modes, where the optical mode is plotted for the four levels of turbulence, thus four levels of absorption and scattering.We modeled all modes by assuming transmission power control with up to 10 W. Note that 0.1 W power is always needed for the mode operation.The 10 W transmission power Fig. 9. Propagation delay and energy distribution for 51-nodes randomly located along a chain.imposes a limit on the maximum transmission range.For the acoustic mode this range is high due to the lower spreading factor of the acoustic propagation.For the optical mode, the Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.range highly depends on the absorption and scattering: this component is exponential, thus it leads to a relatively hard limit on the transmission distance.This optical communication range can be regarded as underwater visibility distance, since the wavelength range that is optimal for the underwater communication lies within violet to green colors of the visible light spectrum.
While the graphs of the required transmission power show the achievable ranges, they cannot really be used as a good metric for communication decisions because of very different communication rates.The normalized qualities of the required energy per bit versus distance is plotted on Fig. 5. From Fig. 5, we can clearly see that the optical mode has the best energy efficiency in most scenarios.The only exception is the case of a very turbid water with a very short visibility, where the MI mode provides better efficiency at distances of several meters.In reality, the MI mode would be advantageous in other cases, e.g. when the optical modes do not have line-of-sight visibility, or the background illumination is high.
Thus, we can clearly see that, on the one hand, both optical and MI are good for the short-range communication, and, on the other hand, the acoustic communication should be left for long-range hops where neither optical nor MI can provide connectivity.

B. Numerical Analysis
In this subsection, we provide numerical evaluation for the performance of the multimode UniSDM systems over several network topologies.The performance is assessed based on two metrics: the transmission delay, and the energy required for the transmission.We start with a trivial example of a chain topology with equidistant node spacing, to provide useful insights on the UniSDM system performance.We then switch to the topologies with randomized nodes' locations: a chain with random nodes' locations, and a square area with random nodes' locations.These scenarios are modes of networks comprised by fleets of underwater vehicles, assuming we have no control on the vehicles movement.
We start with the two configurations: • multimode nodes with acoustic and optical modes • acoustic-only nodes as a baseline case.
We simulate n = 1000 bit packet transmission.This packet size leads to a packet transmission time t tx ,m = n/b m ms, and m ∈ {ac, opt, MI , RF }.
This leads to the transmission times of t tx ,opt = 0.1 ms for the optical mode, and t tx ,ac = 20 ms for the acoustic mode just due to the modes' transmission rate.The total hop time is the sum of the transmission time and the propagation delay, where d hop is the hop length.
The maximum transmission power is limited to P max = 10 W for any mode.For each hop, the minimal required power P tx is computed, such that a reliable connection is established.To make the power control simulation more realistic, we added P b = 0.1 W base power consumption for a mode operating at Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.minimal power, and thus the total energy consumed per hop is 1) Chain With Uniform Locations: Let us consider a chain of 151 nodes with equidistant spacing of 10 m along a total length of 1500 m, as shown on Fig. 6.It is trivial to compare propagation-delay minimizing scenarios for both cases, acoustic only and multimode.
Acoustic-only stations can transmit signals from the first station to the last in 1 hop and this is taking 1.02 s.The data rate is 50 kbps.
Multi-mode stations can use their optical modes.If we assume 150 hops transmission, the transmission time is 15ms (ignoring possible interferences), and the propagation delay is negligible.The capacity would be a fraction of 10 Mbps due to interference, that is still orders of magnitude more than the acoustic.The actual transmission would use longer hops, since it is possible to transmit data in 30 to 50 m directly.This example clearly shows that the optical transmission outperforms acoustic transmission in terms of delay and capacity.It is easy to show that it is also much more efficient in terms of energy consumption per bit per meter of transmission distance.
It is important to note that the number of hops is not an adequate metric for this kind of systems with highly heterogeneous modes: the route with 150 hops clearly outperforms 1-hop route.
2) Chain With Random Locations: To show more realistic application of multimode modems, we simulate two scenarios, with N = 151 and with 51 stations randomly distributed along 1.5 km path, respectively.The topology is shown on Fig. 7.
We implemented a simulation that computed the most efficient path according to one of the two metrics: the total packet transmission delay, or total packet forwarding energy.To compute the packet delay, per-hop delays and energies are computed according to Eq. 5 and 6 for all the hops of a route and summed up. 100 random instances of the scenario were evaluated to calculate the cumulative distribution functions (CDFs).
The CDF of the total transmission delay and total packet forwarding energy are shown in Fig. 8 for 151-nodes chain.From the figure, the system where only acoustic link is Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
used always needs around one second and 0.07 J per packet transmission.One the other hand, multimode (acoustic-optical system) the performance varies depending on the nodes distribution.As can be seen in Fig. 8 (top panel), the delay is less than 0.003 s with probability higher than 80% and less than 0.2 almost always.For the energy consumption of the acoustic-optical system, Fig. 8 (bottom panel) shows that the consumed energy is less than 0.002 J with probability of 80% and almost always less than 0.01 J.
For the 151-nodes chain scenario, the sudden increase in transmission delay and energy consumption at the CDF value of 0.8 is due to the need for at least one acoustic hop.This is due to the fact that, while the average distance between the nodes equals 10 m, it also considerably varies and we observe a 80% probability of a hop longer than ≈60 m that is the optical communication limit.
In other words, a system that only relies on optical communications would fail in this scenario with a probability of around 0.2, and therefore an acoustic backup link is needed.The steps that we observe in the graphs correspond to adding one acoustic hop or two acoustic hops.With no acoustic hops we see ∼1000 times delay gain and ∼ 100 times energy consumption gain, compared to the baseline case with only acoustic communication (red line).In presence of acoustic hops they drop to ∼10 times, compared to the baseline case.
The scenario where 51-nodes chain is considered (instead of 151 nodes) is shown in Fig. 9.When we simulate 51-nodes chain instances, the average hop length equals 30 m, but there are always some hops longer than ≈60 m.For this reason, a system that only relies on optical communications would not be able to operate in these conditions, and a multi-modal system always uses the acoustic mode for some of the hops.Because of that, we do not see the cases of 100 times delay or energy consumption improvements that correspond to opticalonly propagation.Though we can observe significant gains of 1.5 to 5 times for delay, and 2 to 5 times for the energy consumption gain, since most of the path is still covered by the optical mode.
3) Square Area With Random Locations: Fig. 10 shows a scenario of data transmission from node number zero at the lower left corner to node number one at the upper right corner of the square area.The source and destination nodes are fixed at these locations, all the other nodes are placed according to uniform random coordinate distribution.Blue edges show established optical communication links, and green edges show acoustic communication links.It can be seen that the lowest-delay route uses optical links as much as possible.Because of limited optical communication links, there are no optical connectivity to neither node #1 nor node #0, thus the first and the last hops are acoustic.
The simulation results are shown in Fig. 11.We can clearly see that the general results are the similar to the results for a linear topology.We observe a 90% probability of a pure optical transmission that provides several orders of magnitude gain, and some improvement even for worst scenarios.
We can also see that in this case, characterized by fair visibility conditions (coastal ocean turbidity, no significant background light, line of sight visibility), the magnetic induction mode does not provide any significant benefits over the optical mode in terms of transmission time or energy consumption.Thus, the curve for Acoustic-MI case coincides with Acoustic, and Ac.-Opt.-MIcoincides with Acoustic-Optical case.
4) High Turbidity Scenario: As it was shown in the discussion of the particular modes, the magnetic induction mode achieves benefits over the optical mode only under some conditions: very turbid water, no LoS visibility or interference from the background light.To show the operation of the MI mode in our model, we created a scenario with a very turbid water (with absorption and scattering coefficient of 10 m −1 ), and relatively dense nodes' placement.The results are shown in the Fig. 12.
We can clearly see that the "MI plus acoustic" configuration gives a significant advantage over acoustic-only communications.This case is similar to adding an optical mode in clear water: the lower delay and higher bandwidth mode has limited range, thus, it is used for most of the route except for the longer hops that require the use of acoustic mode.
Due to a very limited optical communication range, adding an optical mode does not benefit this scenario much.Some minor improvements can be seen on the delay plot.
The energy plot shows no significant gain since MI energy efficiency is comparable to short-range acoustic communication.Adding the optical mode gives no significant gain since most of the energy consumption is defined by the MI and acoustic modes.Because of that, the curves for the Acoustic case coincide with Acoustic-Optical, and Acoustic-MI case coincides with Ac.-Opt.-MIcase.

V. CONCLUSION
In this paper, we have proposed the architecture of the universal multimode modem, UniSDM.This modem allows joint and seamless operation of acoustic, optical, magnetic induction and RF modes.Furthermore, the implementation of various technologies at the same time is possible, due to the modem software-defined architecture.We provide details for the UniSDM operation and discuss efficient multi-mode cooperation.In comparison with traditional setups with separate modems, the UniSDM improves synchronization and transmission rate, and reduces latency, all at reduced transceiver cost and size.
Compared to our previous conference paper [1], in this work we introduced an algorithm for the modes' cooperative use.We also added a discussion of the modes' features, showing their strong and weak sides and how they complement each other.We estimated the potential UniSDM performance by doing a numerical evaluation.In addition to a comparison of pure acoustic and dual-mode acoustic and optical configurations, we evaluated the configurations with the magnetic induction mode into consideration.We also added a square area scenario with random nodes' location, which showed the results similar to a chain topology.
The numerical evaluation showed that the multi-modal approach achieves massive performance improvements in many scenarios.Furthermore, since it reduces the usage of acoustic transmissions, it improves acoustic interference situation, and also provides less impact on the marine life.
We showed that the multi-modal underwater communications have unique features of combining highly heterogeneous modes.It also provides some specific challenges.For example, a lowest-delay route can include multiple non-acoustic hops, thus the number of hops can no longer be used as a good routing metric.Thus, future research on the multi-mode underwater communications, such as media access and routing protocol design, would have to count for these features.
Along with the universal access to the communication modes and technologies that use them, the software defined architecture would facilitate additional applications, such as navigation and monitoring.These problems can also be addressed in the multi-modal paradigm, e.g. the navigation may combine radio navigation (including the use of global navigation satellite systems) and underwater acoustic navigation (including long baseline and ultra-short baseline positioning) approaches.The UniSDM can be used as well to monitor the aerial RF channel and to track underwater acoustic channel at the same time.
In the future, we plan to show the UniSDR operation on the hardware.Currently, we are working on the hardware design and the underwater communication testbed, as required by our internal projects based on AUVs' fleet.

•
active communication • standby (active RX) mode • sleep (i.e.requiring external request to power-up) • hibernated, i.e. not loaded to DSP and needing additional time and possibly reboot for enabling • unavailable

Fig. 10 .
Fig. 10.A scenario with random locations of 51 nodes in a 200 × 200m area.Water model of coastal ocean.

Fig. 12 .
Fig. 12. Propagation delay and energy distribution for 21-nodes' chain of 50 m length.High water turbidity case.
Power level control, i.e. controlled amplification.The electric interface between SU and MF use standardized signal levels for the interconnection.On the other side, signals' power are adjusted before being converted to another medium for transmission.Conversely, the received signals' powers are adjusted to standardized levels before being processed at the SU and the BBU.In order to match the signal power levels, the interface between SU and MF shall carry information on the required transmission signal power, and average received signal power.• Conversion between baseband electric signals and signals in actual physical modes, i.e. acoustic, RF, MI, and optical signals.The signals' conversion for each of the four communication modes are here described: • Control communication.Along with the BBU control signals, control signals can be generated and communicated at the SU, to provide information on the mode of operation selected for a MF.• Synchronization and Timing (S&T).•(Optional) Frequency up-/down-conversion.For high frequencies (RF and beyond) the baseband signal generated by SU must be converted to actual radio frequency.The electrical-optical conversion may be viewed as a variant of frequency conversion, e.g. a transmission at 400 nm wavelength essentially is a transmission at 750 THz carrier frequency.While current research is focused on (offset) amplitude modulation techniques, future underwater optical communication might use coherent optical modulation schemes.• -Radio Frequency MF (RF-MF) have ubiquitous use, and are well developed.Basically they consist of local oscillators (LOs), mixers, filters and amplifiers.To provide capability of MIMO operation using multiple RF-MFs, the LOs must be synchronized.Thus, the synchronization connections come from the SU.Simplified architecture of the RF-MF is shown in Fig. 2. -Acoustic MF may have widely varying frequencies.In general, transmission frequencies are low compared to RF, hence, additional frequency conversion is not needed.Simplified architecture of the acoustic MF is shown in Fig. 3. Depending on frequencies and dynamic range requirements, the receivers (hydrophones) and transmitters (transducers) may be combined or separated.-Optical MF presents a wide variety of options.Unlike other modes, optical Rx and Tx components are always separate.The transmitter side may be implemented by LEDs (light-emitting diodes), LASERs or other means.