Multi-Level Multi-Technology Underwater Networks: Challenges and Opportunities for Marine Monitoring

Abstract

Underwater networks are crucial for monitoring the marine ecosystem, enabling data collection to support the preservation and protection of natural resources. Among the various technologies available, acoustic and optical communications stand out for their superior performance in underwater environments. Acoustic technologies are suitable for long-range communications, typically operating over hundreds of meters up to several kilometers, albeit with low data rates ranging from a few hundred bps to few tens of kbps. In contrast, optical technologies excel in providing high data rates, often between 1 and 10 Mbps, but only over short distances (e.g., 50 m) in controlled conditions. To leverage the strengths of these technologies, recent research has proposed multi-modal underwater systems; however, these solutions generally rely on single-level or at most dual-level architectures, limiting the benefits of a structured hierarchical approach. In this review paper, after discussing related work on multi-technology acoustic and optical networks, we highlight relevant design guidelines for multi-technology, multi-level underwater architectures, explicitly considering three layers: a deep acoustic layer, an intermediate optical layer, and an upper RF-enabled surface layer. For illustration, we also discuss a PoC of such a hierarchical architecture under development at the University of Catania, Italy, in the Area Marina Isole dei Ciclopi natural reserve. The PoC includes optical nodes capable of transmitting up to 10 Mbps over short ranges and acoustic nodes (both software defined and not) supporting rates of tens of kbps over hundreds of meters and being adaptive to network conditions, interconnected through hybrid multi-technology nodes deployed across the three network levels. By assigning specific technologies to appropriate layers, the architecture enhances scalability, robustness, and adaptability to dynamic underwater conditions. This design strategy not only improves data transmission efficiency but also ensures seamless operation across diverse marine scenarios, making it an effective solution for a wide range of underwater monitoring applications.

1. Introduction

As oceans cover nearly 71% of Earth’s surface, they are fundamental to regulating the climate, sustaining biodiversity, and supporting human livelihoods. Similarly, underwater networks play an essential role in modern communications, as submarine cables carry 99% of global telecommunications traffic, connecting equipment across oceans and enabling seamless data exchange [1]. In the post-COVID era, with an unprecedented reliance on digital services, the demand for these underwater networks has surged, driving major investments from technological companies like Google and Microsoft, which have collectively spent over 10 billion dollars on new subsea cables [2]. This demand is expected to grow at a compound annual rate of 50% between 2024 and 2028, reflecting the increasing need for high-capacity resilient communication systems.

Beyond traditional submarine cable networks, advances in underwater wireless sensing and communications are particularly critical for addressing unique challenges related to security in marine environments. Unlike terrestrial and space communications, underwater systems face significant limitations, such as signal attenuation, limited bandwidth, and energy constraints [3].

The oceans vastness and complexity require innovative communication systems capable of overcoming extreme depth, marine currents, and high-pressure environments. UnderWater Sensor Networks (UWSNs) have emerged as pivotal solutions for efficient and automated data acquisition. UWSNs enable a wide range of applications, from monitoring climate change impacts on marine ecosystems and assessing water quality to disaster response and industrial operations like oil exploration [4]. However, effective implementation of these networks requires overcoming the fundamental limitations of underwater communication scenarios.

Acoustic communications (ACs) are a mature technology, widely employed for long-range underwater transmissions, capable of covering distances up to tens of kilometers. They offer reliability in harsh underwater conditions but are constrained by low data rate, significant propagation delay, and vulnerability to environmental noise [5]. On the other hand, optical communications (OCs) provide higher data rate and faster propagation speed, making this ideal for short to intermediate communication ranges (up to hundreds of meters). However, OCs require clear water conditions and precise alignment due to the sensitivity to turbidity and line-of-sight constraints [6]. These complementary characteristics underscore the need for hybrid multi-technology systems that combine the potentials of AC and OC while mitigating their respective weaknesses.

As an alternative to the above approaches, the use of radio frequency (RF) waves for underwater applications unveils the possibility to achieve very high transmission rates, although in the proximity of water surface only. Indeed, radio frequency exhibits a short-range high data rate near the water surface and offers seamless integration with terrestrial networks. However, key challenges in terms of high attenuation met due to marine water conductivity and permittivity still remain [7]. Energy inefficiency is another constraint because of the required transmission power in the order of Watt and the large antenna needed. Accordingly, in the rest of this paper, we consider the use of opto-acoustic communications only for the underwater part, while we will explore the possibility to employ RF waves for onshore–offshore–cloud data delivery and network management only.

The integration of different technologies into unified multi-level systems holds the potential to revolutionize underwater communications. Hybrid hierarchical systems can balance range, speed, and energy efficiency tradeoff, offering robust performance across diverse underwater conditions. Recent research highlights the technical efforts required to synchronize these modalities, ensure seamless data exchange, and address challenges such as biofouling and corrosion, as well as consequent hardware durability. By leveraging the complementary capabilities of AC, OC, and RF, multi-technology systems can enhance the efficiency and resilience of underwater networks, paving the way for cutting-edge applications in marine environments.

This work, provides an initial excursus across different multi-technology underwater networks (AC or OC) and then specifically focuses on the possibility to integrate a subset of the technologies to set up multi-technology, multi-level underwater communication systems, addressing challenges and issues in their design and implementation, and confining their limitations. By exploring these complexities, we aim to advance the development of efficient and scalable multi-technology systems for underwater communications. Also, a PoC of such a multi-level, multi-technology architecture under development at the University of Catania is illustrated.

The rest of this paper is organized as follows. In Section 2, we report some of the relevant literature on opto-acoustic UW networks and software-defined acoustic nodes. In Section 3, strengths and weaknesses of these technologies are detailed separately. In Section 4, guidelines for design of a multi-level, multi-technology UW architecture are provided, including also an RF edge section to exchange data with offshore platforms and the cloud. In Section 5, an example of a compliant network architecture being developed at the University of Catania, Italy, is described. Finally, open issues and challenges are discussed in Section 6.

2. Related Work

2.1. Opto-Acoustic Multi-Technology UW Networks

Acoustic communications are well suited for underwater environments due to their extended transmission range and enhanced reliability, which are crucial in such challenging contexts. Conversely, scenarios demanding real-time responses or high data rates benefit from optical communication (OC), although with reduced reliability and coverage range. OC systems excel in applications such as exploring and monitoring underwater ecosystems, particularly when transmitting images and videos over very short distances. Multi-technology communication systems effectively integrate diverse methodologies to meet the numerous requirements of underwater networks.

In the domain of underwater multi-technology communications, Wang et al. [8] proposed a hybrid acoustic–optical communication strategy tailored for underwater ecosystem exploration and monitoring. Their methodology employed OC for short-range, high-data-rate transmissions and AC for disseminating control and localization messages. To extend the range of OC, they utilized multi-hop communications, space division multiple access (SDMA) at the data link layer, and a reverse route search strategy at the network layer. Furthermore, they incorporated SNR-based adaptive switching to switch between communication modes. While this mechanism optimizes performance under varying channel conditions, it also introduces certain drawbacks. Frequent switching between communication modes, particularly in dynamic underwater environments, can result in additional energy consumption and transmission delay. Each mode transition requires recalibrating beams, reconfiguring protocols, and updating routing paths, creating overheads that may undermine the system energy efficiency and timeliness in highly variable scenarios. Also, exchange of much information is needed to let the system dynamically perform design choices with the drawback related to the dynamic channel variations and the excessive overhead, which in an unreliable environment like the marine one can become unbearable.

Similarly, Eduardo et al. [9] introduced an opto-acoustic hybrid underwater data collection framework in a simulation-based study focusing on energy efficiency and scalability. The protocol employed a clustering approach, selecting cluster heads with the highest-scoring node based on energy and connectivity within the cluster. Intra-cluster communication utilizes optical links for energy-efficient and high-speed data transfer. These clusterheads aggregate data from member nodes and transmit it to the sink node by employing AC in inter-cluster communications. They also used a time-division multiple access (TDMA) scheduling mechanism within clusters to ensure collision-free communications, while a routing tree protocol guides inter-cluster data transmissions to the sink node. Simulation results indicate that this clustered multimodal approach performed better as compared to the shortest path algorithm in terms of energy consumption, latency, number of transmissions, and throughput. Also, the use of AC for inter-cluster communication consumes significantly more energy than the optical nodes, resulting in a reduction in overall network lifetime.

Han et al. [10], in their simulation-based study, demonstrated the superior performance of hybrid opto-acoustic communication when compared to standalone acoustic and optical modes in terms of throughput. Using constant bit rate (CBR) traffic, they configured packet sizes at 1.75 kB for AC and 50 kB for OC. In their system model, the nodes were equipped with both optical and acoustic modems. The AC is used for long-range transmissions and omnidirectional links to establish network topology between nodes. Along with this, the localization and optical node alignment was also performed by these AC links. While the study highlights the potential of the hybrid opto-acoustic communication system, it overlooks critical environmental factors, such as water temperature and turbidity, which heavily influence underwater communication reliability. Additionally, the research has not explored adaptive or hierarchical approaches that could enhance the system flexibility and performance under dynamic conditions.

Islam et al. [11] developed an analytical framework to calculate the transmission power requirements for acoustic and optical underwater communication systems. Their aim was to optimize the node power consumption when transmitting data from underwater nodes to surface stations. Their proposed system is structured assuming that the member nodes were designated to collect data, clusterheads perform aggregation and forwarding, and relay nodes are used to form multi-hop links. They use OC for short-range links while AC for longer range where adverse environmental conditions can occur. Simulation results revealed significant power savings of 23% as compared to purely acoustic or optical approaches. Despite highlighting the power-saving advantage, the study acknowledges potential challenges in solving the optimization process, though effective but computationally intensive, and highlights scalability issues in case of larger deployments. Alongside this, the OC is significantly affected by environmental conditions, like water turbidity and light attenuation, which can compromise reliability in extreme scenarios. Additionally, there is a need for heuristic approaches to simplify the computational methods and improve scalability, especially in larger and more dynamic network environments.

In a practical demonstration, Vasilescu et al. [12] designed an underwater opto-acoustic multi-technology communication system for marine ecosystem monitoring. Their design used AC for broadcast messages at 330 bps over distances exceeding 400 m, while OC enabled data transmission at 320 kbps over a range of 2 m. Similarly in another study [13], the authors proposed a hybrid opto-acoustic communication system to enable real-time underwater video streaming. The system integrates optical modems for high-speed, short-range video transmission and acoustic modems for long-range communications. Acoustic signals are used for localization, modem alignment, control messages, and network topology formation. Additionally, in an experimental testbed, they used AquaSeNT acoustic modems in a water tank to validate the feasibility of compressed image transmission. They compared different image compression techniques to optimize bandwidth for low-quality video transmission over acoustic channels when the environmental conditions are not optimal. Note that the image compression techniques effectively reduce the image size but add an extra computational complexity on the underwater energy-limited nodes. Additionally, the lack of an adaptive protocol approach limits the adaptability and scalability in large-scale underwater networks.

Qiontas et al. [14] developed a hybrid opto-acoustic underwater network, evaluating its performance through simulations and testbed experiments. Their approach integrates high-speed OC for short-range data transmission with long-range AC for coordination and control messages. They also use a modular cognitive communication architecture (CCA) to enable the adaptive selection of protocols based on environmental and communication conditions. In simulations, they achieved approximately a 95% packet delivery ratio (PDR). They validated the system through real-world experiments with a Medusa-class vehicle by considering four scenarios. The optical-only and acoustic-only scenarios achieved PDRs of 100% and 97.4%, respectively, at data rates of 10.5 kbps and 842 bps. However, the hybrid approach exhibited decreased performance during cooperative path forwarding as compared to the fixed-path optical scenario. This degradation was attributed to the absence of line-of-sight links, resulting in the frequent use of acoustic communication, which significantly reduced overall efficiency.

Bhatnagar et al. [15] explored a dual-hop hybrid underwater system integrating acoustic and optical links to monitor shallow-sea environments. The paper presented an underwater opto-acoustic sensor network, where acoustic and optical sensor nodes communicated with an underwater vehicle relay through decode-and-forward relaying. This method aimed to achieve higher data rates and reduced latency versus systems using only acoustic links.

Following a similar approach, He et al. [16] presented a cooperative communication platform for underwater environments. This platform was designed for a remotely operated vehicle (ROV) with LEDs and acoustic modems. The system used optical communication for large data transmission, whereas acoustic communication enabled signal exchange before optical transmission and supported automatic repeat requests (ARQs). The platform achieved a data rate of up to 5 Mb/s over a 7.6 m laboratory testing, with an estimated potential link distance of 11 m in clear seawater at a rate of 3.125 MB/s.

Subsequent contributions extended hybrid acoustic–optical communication to AUV-assisted data collection and scheduling in underwater sensor networks. Bu et al. [17] presented an AUV-assisted optical–acoustic hybrid data collection framework optimized using deep reinforcement learning (DRL). This framework selected the communication modality based on data significance and packet size while minimizing both the age of information (AoI) and energy consumption. The scheme employed a multi-modal steering angle optimization algorithm to reduce the AUV navigation overhead. Nonetheless, the model simplified optical communication by omitting the alignment time and assumed omnidirectional optical modems, which may have led to overestimation of the optical performance in turbid water conditions.

In parallel, Zhang et al. [18] presented a hybrid opto-acoustic framework designed for the localization and tracking of AUVs. This framework employed acoustic signals to estimate the angle of arrival for position feedback and utilized adaptive model-predictive control to maintain the AUV within the optical communication conical range. The study demonstrated that adaptive control reduced tracking errors by 53%. However, the design did not consider the stability of the 3D optical link. Furthermore, the optical model was based on exponential optical attenuation, which was applicable only to clear seawater where the scattering effects are minimal.

More recent work further advanced hybrid systems toward higher data rates and closer acoustic–optical coupling through interactive designs. Fan et al. [19] developed an underwater optical–acoustic interactive system capable of achieving nearly 100 Mbps over a distance of 22 m in water tank experiments. The system utilized acoustic feedback to adaptively control optical parameters for alignment and error correction. It demonstrated seamless integration with Ethernet, achieving a bit error rate (BER) of ≤10⁻⁷ at data rates up to 10 Mbps. However, the validation of this system was confined to laboratory conditions, and its performance in turbid coastal waters has not yet been characterized.

Recent advancements in opto-acoustic integration have extended beyond link-level control and localization. Zhang et al. [20] introduced the AO-CLOR, a cross-layer routing protocol for underwater acoustic–optical hybrid sensor networks. The protocol used omnidirectional acoustic links to disseminate control and position information, whereas short-range optical links transmitted high-rate data. Simulations showed that this approach enhanced the data rate and packet delivery compared to single-modality designs. Gorodilov et al. [21] developed a UniSDM proof-of-concept multimode software-defined modem that integrates acoustic, optical, magnetic induction, and RF frontends within a flexible firmware stack supporting JANUS and OFDM standards.

In

Table 1. Comparison table for related work.

Study	Methodology	Key Results	Strengths	Limitations
2017 [8]	Hybrid opto-acoustic system with multi-hop OC, SDMA for data link layer, and reverse route search.	Effective for short-range, high-data-rate OC and long-range AC for control/localization messages.	SNR-based adaptive switching; scalable multi-level design.	Frequent mode switching increases energy and delay; limited real-world testing.
2020 [9]	Clustering-based hybrid framework using optical intra-cluster and acoustic inter-cluster communications.	Achieves 73% energy savings and reduced latency as compared to shortest path routing.	Energy-efficient clustering, scalability, high throughput.	Energy-intensive inter-cluster AC links; high latency in dense networks.
2014 [10]	Simulation study using CBR traffic to compare standalone vs. hybrid opto-acoustic systems.	Hybrid mode achieves higher throughput as compared to standalone systems.	Proves hybrid performance; realistic packet configurations for OC and AC.	Environmental factors, like turbidity, are ignored; lacks adaptability to dynamic conditions.
2021 [11]	Hierarchical structure for OC (short-range) and AC (long-range) with power optimization framework.	Achieves 23% power savings as compared to standalone systems.	Efficient power management.	Computationally intensive optimization; scalability issues in complex networks.
2007 [12]	Practical multimodal design with OC for data and AC for control/localization.	OC: 320 kbps over 2 m; AC: 330 bps over 400 m.	Real-world validation.	Limited OC range; AC low data rate unsuitable for data demanding applications.
2019 [13]	Hybrid system for real-time video streaming using OC for video and AC for control/alignment.	Validates real-time streaming with compressed image transmission.	Proves feasibility of underwater video; effective compression techniques.	Computational overhead; lacks adaptability for larger-scale networks.
2021 [14]	Modular cognitive communication architecture (CCA) for hybrid systems with adaptive protocol selection.	Achieves 95% PDR in simulations; validated with Medusa-class vehicles.	Adaptive protocol improves efficiency; real-world validation with vehicles.	Line-of-sight absence reduces efficiency; frequent reliance on AC affects hybrid performance.
2023 [15]	Hybrid dual-hop DF relay with $\alpha$-F (acoustic) and EGG (optical) models.	Closed-form expressions for outage probability, ABER, ergodic capacity.	Novel physical layer analysis; realistic composite fading models.	Line-of-sight optical vulnerability; static topology; no field validation.
2022 [16]	ROV with LED-based optical and acoustic modems; optical for data transmission, acoustic for signaling and ARQ.	Achieves 5 Mb/s over 7.6 m laboratory pool; estimated 3.125 MB/s over 11 m in clear seawater.	Practical ROV implementation; cost-effective design; experimental validation in controlled environment.	Laboratory-only testing; seawater estimates based on attenuation coefficients, not field-validated; ARQ overhead in lossy conditions.
2023 [17]	DRL-based AUV path planning with adaptive optical–acoustic selection; considers AoI, packet size.	Reduces weighted AoI and energy vs. single modality; demonstrates energy–AoI trade-offs.	DRL approach for multi-modal selection; steering angle optimization.	Omits optical alignment time; assumes omni-directional modems; static network topology.
2020 [18]	Hybrid acoustic–optical; phase-based AoA localization; adaptive MPC and PD controllers; 2D control.	MPC 53% better than PD; maintains QoS within cone.	Position-error feedback; unknown dynamics handling.	2D control (no 3D attitude); perfect alignment assumption; clear-water model.
2025 [19]	Optical–acoustic hybrid with adaptive parameter control via acoustic feedback.	100 Mbps at 22 m (lab); 66 m equivalent range in Jerlov I water; BER $\le {10}^{-7}$.	Practical integration; seamless Ethernet compatibility; adaptive alignment.	Lab-only; no ocean trials; turbidity effects unknown.
2024 [20]	Acoustic links for control/position broadcast, optical links for DATA forwarding via sector-based candidate areas.	Simulations in a 500 m × 500 m network (60–150 nodes) show higher delivery ratio, lower delay, and less energy consumption as compared to acoustic or optical links.	Exploits complementary acoustic/optical properties; uses local information (distance, residual energy, neighbor count) for distributed next-hop selection.	Evaluation is simulation-only in 2D; idealized localization; no experimental or 3D mobility validation.
2025 [21]	Multi-mode SDR modem with acoustic, optical, MI, and RF heads.	Lab PoC with 3-node setup showing acoustic–optical relaying, Ethernet tunneling, video streaming, and multi-mode test tools.	Highly flexible architecture; supports multiple media and protocols.	Demo-scale only; real-environment performance not evaluated.

, we summarize the main contributions described above, detailing the key results, strengths, and weaknesses.

Although previous studies have significantly contributed to opto-acoustic underwater communications, our work discusses a multi-level multi-technology (or multi-modal) architectural framework aimed at improving adaptability and efficiency in underwater communications. In this architectural framework, acoustic and optical communication levels are combined while also including the possibility to exploit software-defined acoustic modems to enable real-time tuning of various parameters at the acoustic level, thereby enhancing the reliability and performance of the communication system under highly dynamic environmental conditions. Moreover, a surface level including RF-enabled nodes is proposed to create a modular and scalable system for various underwater applications where data can be accessible remotely and actuation can be also performed on underwater devices.

2.2. Software-Defined Acoustic Nodes

Building on the evolution of software-defined underwater acoustic modems reviewed in [22], the concept of software-defined acoustic nodes (SDANs) represents a natural and necessary architectural progression from isolated, flexible modems toward fully programmable, network-aware underwater communication systems. While early software-defined acoustic modems primarily focused on increasing reconfigurability at the physical layer, recent advances in embedded processing, operating systems, and middleware have enabled a more holistic approach in which sensing, communication, networking, and control functions are jointly considered within a unified node architecture. In this context, SDANs can be viewed as underwater counterparts of terrestrial Software-Defined Networking (SDN) systems, adapted to the unique constraints of the acoustic medium [23].

At the architectural level, SDANs embrace the principle of logical separation between the data plane and the control plane, a cornerstone of SDN architectures in radio and wired networks [24]. The data plane encompasses all time-critical operations required to transmit, receive, and forward acoustic data. These include waveform generation, modulation and coding, synchronization, equalization, Doppler compensation, channel estimation, medium access control, and packet forwarding. In contrast, the control plane is responsible for higher-level decision-making processes such as protocol selection, parameter adaptation, routing policy enforcement, resource allocation, and mission-driven coordination among nodes. This separation enables independent evolution and optimization of control logic and data processing, thereby improving flexibility, maintainability, and interoperability.

In SDAN architectures, the data plane is typically implemented using software-defined modem technologies running on general-purpose processors or heterogeneous computing platforms combining CPUs with DSPs or FPGAs. As demonstrated by systems such as NILUS SoftModem [22], UnetStack-based modems [25], and GNU Radio–based acoustic platforms, modern embedded processors are increasingly capable of executing complex physical-layer algorithms in real time while also supporting full network protocol stacks. By implementing the physical and link layers in software, SDANs allow acoustic waveforms, coding schemes, symbol rates, bandwidth usage, and MAC behaviors to be dynamically reconfigured at runtime. This capability is particularly valuable in underwater environments, where channel conditions can change rapidly due to mobility, multipath, and environmental factors, and where preconfigured static protocols often result in inefficient or unreliable communication.

The control plane operates on longer time scales and maintains a broader view of the network state. Depending on the deployment scenario, it may be centralized at a surface gateway, distributed across underwater nodes, or organized hierarchically to balance responsiveness and robustness. Centralized control can simplify network optimization and policy enforcement but may suffer from latency and reliability issues due to intermittent acoustic links. Distributed or hierarchical control, on the other hand, enables localized decision making while still supporting coordinated behavior when connectivity permits. In all cases, the control plane leverages network state information such as link quality indicators, queue occupancy, residual energy levels, node roles, and application priorities to guide adaptation decisions [23].

A critical aspect of SDAN design is the definition of clear and standardized interfaces between control and data planes. Conceptually analogous to southbound interfaces in SDN, these interfaces allow control logic to configure data-plane parameters, such as transmission power, carrier frequency, modulation scheme, coding rate, retransmission policy, and sleep schedules. Unlike terrestrial SDN protocols, however, underwater southbound interfaces must account for long delays, low data rates, and high packet loss, which often preclude frequent fine-grained control updates. As a result, SDANs typically rely on coarse-grained configuration commands, policy-based control, or predictive adaptation mechanisms rather than continuous feedback loops.

Complementing the southbound interfaces, SDAN architectures also define northbound interfaces that expose abstract communication and sensing services to higher-layer applications. These interfaces allow applications to specify requirements in terms of latency, reliability, data freshness, or energy consumption without being tightly coupled to specific physical-layer implementations. For example, an application may request periodic delivery of sensor measurements with bounded delay, while the underlying SDAN control plane selects appropriate modulation schemes, packet sizes, and routing strategies to satisfy this requirement. Such abstraction facilitates cross-layer optimization and simplifies application development, particularly in heterogeneous networks composed of nodes with different hardware capabilities and roles.

Interoperability and extensibility are further enhanced through the use of open and modular software interfaces within the node itself. Many SDAN implementations adopt message-oriented middleware, publish–subscribe mechanisms, or socket-based APIs to decouple physical-layer processing from MAC, network, and application layers. Frameworks such as DESERT Underwater [26,27,28,29] exemplify this approach by enabling the same network protocol code to be used in simulation, emulation, and real-world deployments. Similarly, agent-based architectures like UnetStack [25] provide modular services that can be dynamically instantiated, configured, or replaced at runtime. This modularity not only accelerates experimentation and prototyping but also supports long-term maintainability and evolution of deployed systems.

From a system-level perspective, SDANs enable advanced networking functionalities that are difficult or impossible to realize with traditional monolithic acoustic nodes. These include policy-driven protocol switching, cognitive adaptation to environmental conditions, cooperative resource management, and integration with digital-twin and mission-planning frameworks. For instance, a control-plane entity may instruct a subset of nodes to switch to a robust low-rate waveform during periods of high noise, while others maintain higher-rate links where channel conditions permit. Similarly, SDANs can support seamless transitions between different operational modes, such as discovery, data collection, and emergency signaling, without requiring node reboot or physical intervention.

Despite their advantages, SDAN architectures also introduce challenges, particularly in terms of software complexity, energy consumption, and real-time performance guarantees. Running full operating systems and flexible protocol stacks on embedded platforms typically increases idle power consumption and wake-up latency compared to highly specialized hardware solutions. Moreover, ensuring deterministic behavior for time-critical physical-layer processing can be challenging in the presence of multi-tasking and operating system overhead. These trade-offs must be carefully considered in the context of the target application and deployment scenario. Nevertheless, ongoing advances in low-power embedded processors, real-time operating systems, and energy-aware scheduling techniques are steadily reducing these limitations.

In summary, software-defined acoustic nodes represent a unifying architectural paradigm that brings SDN-inspired principles to underwater acoustic networks. By decoupling control and data planes, exposing well-defined interfaces, and leveraging software-defined modem technologies, SDANs provide the flexibility required to cope with the harsh and variable underwater environment while enabling interoperability, adaptability, and long-term system evolution. As underwater applications increasingly demand autonomous operation, heterogeneous cooperation, and tight integration between communication, sensing, and control, SDANs are poised to play a central role in next-generation underwater sensor networks and autonomous maritime systems.

3. Strengths and Weaknesses of Optical vs. Acoustic Technologies for UW Communications

In this section, we specifically discuss the unique strengths that both the optical and acoustic technologies show at two different hierarchical functioning layers. This in order to understand the reasons why optical technology can be proficiently proposed to be employed for the higher network level and acoustic technology for the lower one.

3.1. Optical Technology

Underwater Optical Communication (UOC) systems offer distinct advantages over traditional acoustic communication (UAC) systems used in underwater environments. These benefits make UOC an increasingly preferred choice for high-data-rate and environmentally sensitive applications. More specifically,

• High Data Rate. UOC systems excel in handling significantly higher data rates as compared to UAC systems. This advantage stems from the broader bandwidth available in the optical spectrum, enabling the rapid transmission of large volumes of data. Optical communications typically utilize blue or green light waves (450 nm to 500 nm), as these wavelengths offer superior penetration in water as compared to red, yellow, or orange wavelengths, which are absorbed quickly. This characteristic makes UOC ideal for data-intensive applications, such as high-definition video streaming and rapid data collection from sensor networks.
• Low Latency Delay. Optical communication systems exhibit significantly lower transmission delay than acoustic systems due to the faster propagation speed of light in water as compared to sound. This low latency makes UOC systems particularly suitable for real-time applications that demand immediate data transmission and quick response times, such as underwater exploration and surveillance.
• Limited Contribution to Marine Noise. Unlike acoustic communication, which could potentially disrupt marine life sensitive to sound, UOC systems rely on light waves that are non-intrusive to marine ecosystems. This environmental-friendly approach is especially advantageous in ecologically sensitive areas, where minimizing interference with marine life is of paramount importance.
• Energy Efficiency. UOC systems combine high data rates and low latency with exceptional energy efficiency. Compared to UAC systems, UOC systems transmit data more efficiently per bit, consuming less power to transmit larger volumes of data. This energy efficiency is crucial in underwater environments, where power resources are often limited, and the operational lifespan of battery-powered devices must be extended.

Despite the strengths offered by UOC systems, their performance is however hindered by several critical challenges, including limited communication range, alignment issues, ambient light interference, and high impact of water clarity conditions. These challenges must be addressed to optimize the effectiveness and reliability of UOC systems. More specifically,

• Limited Communication Range. The communication range of UOC systems is significantly constrained by the inherent properties of light propagation in water. Water molecules and suspended particles absorb and scatter light, leading to signal attenuation over distance. In turbid or rich-of-sediments water, where impurities and turbidity levels are higher, these effects are exacerbated, thus further limiting the communication range. As a result, UOC systems are better suited for short-range, high-data-rate applications rather than long-range communication.
• Alignment and Directivity Issues. Due to the directional nature of light, UOC systems require precise alignment between the transmitter and receiver. Misalignment can result from water current, natural movement of underwater vehicles, or installation in dynamic environments, causing signal loss. While the narrow beam width of optical systems minimizes interference and enhances security, it also increases the complexity of maintaining alignment in the three-dimensional, dynamic underwater environment. Ensuring accurate alignment is technically challenging but critical for the effective operation of UOC devices.
• High Dependancy on Water Clarity. The efficiency of UOC systems is heavily influenced by water clarity. In clear water, light experiences minimal absorption and scattering, enabling higher communication ranges and data transfer rate. However, in turbid or murky waters, elevated concentrations of particulate matter lead to significant scattering and absorption, thus reducing the reliability and consistency of communication. This dependency on water clarity presents a challenge for deploying UOC systems in challenging aquatic environments, ranging from sediment-rich coastal waters to harbor settings.

The ability to support high transmission data rate in controllable conditions makes optical technology the natural choice for the higher levels of the architecture. Indeed, close to the surface, by means of depth buoys, the reciprocal position of the optical equipment can be better controlled so as to allow keeping nodes in a standard cone of coverage, typically in the order of 60° [30], and the clarity of the water can be assessed easily. Also, immunity to coastal noise due to cargos and limited contribution to increasing marine noise in coastal areas where many marine mammals can be present suggest again that choosing this technology for hierarchical aggregation of data and transmission in proximity of water surface to available floating sink nodes can be an appropriate choice.

3.2. Acoustic Technology

UAC systems play a pivotal role in various underwater activities due to their unique characteristics. More in detail,

• Long-Range Communications. Acoustic signals can propagate over long distances underwater, which provides a significant advantage over optical communications that suffer severe attenuation in such an environment. Although sound waves travel at approximately 1500 m/s, a speed much slower than light, they maintain their integrity over extended distances due to their mechanical wave nature [31,32].
• Robustness. UAC exhibits exceptional robustness and reliability in challenging underwater conditions, such as murky and turbid waters where optical signals are severely attenuated. This durability arises from the physical properties of acoustic waves, which propagate through seawater via compression and rarefaction. Consequently, acoustic waves can traverse suspended particles and turbidity with minimal energy loss, highlighting their strength and reliability in harsh underwater environments.
• Flexibility and Scalability. UAC systems are notably flexible and scalable, benefiting from their non-reliance on line-of-sight (LoS) connections. They adapt well to environmental dynamics and variability, including changes in temperature, salinity, and pressure. While these factors can influence system performance, particularly the velocity of sound waves, UAC systems remain highly scalable. Additional nodes, such as underwater sensors, buoys, and AUVs, can be easily integrated into the network in various configurations.

Despite its relevant advantages, UAC faces several challenges due to the harsh properties of underwater environments. These challenges significantly impact the efficiency and effectiveness of communication systems. In particular:

• Limited Bandwidth. UAC systems suffer from restricted bandwidth as compared to RF and optical communication systems. This limitation stems primarily from the absorption of sound in water, which increases with frequency, causing significant attenuation over longer distances. For long-distance UAC, bandwidth is restricted to a few kHz, limiting data rates to a few kbps. Channel bandwidth varies depending on communication distance and bit rate requirements, ranging from a few kHz to a few hundred kHz. Moreover, factors such as scattering, path loss, reverberation, and the exponential absorption of sound further restrict achievable data rates over long distances [33,34].
• Propagation Delay. The speed of sound underwater is substantially slower than the speed of light, resulting in significant propagation delay over long distances. This delay impedes communication synchronization and can cause packet loss, retransmissions, and inefficiencies in real-world applications. Environmental factors such as temperature, salinity, and pressure further influence sound speed. For instance, the average speed of sound waves increases with an increase in temperature in the underwater transmission medium. Variability in these parameters introduces dynamic delay, thus complicating real-time communication [34].
• Noise. Acoustic communication is susceptible to interference from noise, which can originate from both human activities, such as shipping, and natural phenomena, including rain, tides, underwater earthquakes, and aquatic life sounds. Ambient or background noise, such as bubbles, also degrades the received SNR, further complicating reliable communication in underwater channels [35,36].
• Absorption. Attenuation in UAC is primarily caused by the absorption phenomena, where the acoustic energy is converted into heat or absorbed by dissolved substances, suspended particles, or other materials in the water. It further depends on the signal frequency and distance. As the frequency and distance increase [37], the absorption becomes worse, resulting in greater energy loss and a reduction in signal strength.

The discussion above highlights the numerous features of the acoustic technology which make it suitable to be employed at the lower level of the architecture to collect data and send it over long distances, up to the higher network level where the optical technology can be exploited. Indeed, the immunity to water turbidity makes UAC suitable to be employed even in proximity of the seabed while also avoiding generation of marine noise at higher levels, closer to the water surface.

Table 2. Comparison table for underwater communication technologies.

	Optical	Acoustic
Application scenario	clear, seawater, coastal ocean water	deep, shallow water
Communication Distance	tens of meters	kilometers
Propagation Speed	2.25 × 108 m/s	1500 m/s
Data Rate	Gbps	Kbps
Latency	low latency due to high propagation speed	high latency resulting from slower sound speed in water
Energy Efficiency	highly energy efficient approximately 30,000 bits/Joules	highly energy inefficient with only 100 bits/Joules
Transmission Power	few W	tens of W
Frequency Band	545 THz to 667 THz	10 KHz to few hundreds KHz
Factors Impacting on Performance	absorption, turbidity, organic matter	salinity, temperature, pressure
Deployment Complexity	moderate complexity, requires precise alignment	established technology with well-understood deployment practices
Environmental Impact	low environmental impact	potential to disturb marine life due to sound propagation
Benefits	very high data rate, high cost	widely used, long communication range
Drawbacks	affected by scattering, absorption, moderate transmission range, LOS communication	low data rate, shadow zones, very high latency

compares acoustic and optical technologies in terms of different elements. The analysis shows a trade-off between range and throughput. Optical communication is characterized by very high data rates, low latency, and energy efficiency but remains confined to short distances and relatively clear water due to absorption, turbidity, and alignment requirements. In contrast, acoustic communication offers substantially greater coverage and established deployment across diverse underwater conditions, albeit with lower data rates and higher latency.

In

Table 3. Comparison among the different UW technologies in terms of immunity to specific physical parameters.

Parameter	Optical	Acoustic
Conductivity
Permittivity
Absorption
Turbidity
Organic Matter
Salinity
Temperature
Pressure
Frequency
Antenna Design
Alignment
Beam Divergence
Ambient Noise
Multipath Propagation

, we report a qualitative comparison among the degree of impact of multiple physical parameters on the performance of the different technologies. Note that, upon increasing the number of black arrows, an increasingly relevant part is played by the specific parameter. Overall, Table 3 indicates that optical performance is predominantly limited by underwater optical conditions, such as absorption, turbidity, and organic matter presence, as well as by link-geometry factors, including alignment and beam divergence. In contrast, acoustic performance is more susceptible to environmental variability and channel impairments, which encompass salinity, temperature, pressure, ambient noise, and multi-path propagation. Additionally, frequency remains a pertinent factor for both technologies.

In the following section, based on the considerations carried out above, we will detail the design features for the architecture of an efficient multi-level multi-technology UW system.

4. Multi-Level Multi-Technology UW Architecture Design

An underwater sensor network is a distributed set of sensing and communication nodes that monitor key marine environmental parameters. The architecture of the network is organized into three primary levels: two submerged underwater levels and one at the water surface, as illustrated in Figure 1. Each level has distinct features that collectively support efficient data collection and transmission.

The top level of the underwater multi-technology architecture is located at the water surface and serves as the interface between the underwater network and external cloud-based or onshore systems. It typically consists of surface buoys equipped with RF devices—such as LPWAN, mobile, Wi-Fi, or satellite gateways—that receive data from underwater sink nodes and AUVs acting as aggregators and can also relay cloud-generated commands for remote actuation on underwater equipment. Designed to withstand harsh marine environments, these buoys typically feature waterproof enclosures, self-sustaining power systems like solar panels, and may perform preliminary data pre-processing to optimize transmission. Once transmitted, data is stored and processed on cloud servers that provide scalable storage, advanced analytics, and remote accessibility through web or mobile applications, enabling researchers and stakeholders to derive actionable insights for monitoring and management tasks.

The intermediate level employs optical modems capable of supporting high-speed data transmission over limited distances. Optical nodes are strategically deployed in regions of interest, such as the epipelagic zone, where monitoring is essential due to its biological significance. This zone hosts most oceanic organisms due to photosynthesis and is critical for studying coral reefs, ecosystems, and tracking cetaceans and other marine species. For instance, underwater optical cameras capture high-resolution images and videos of coral reefs, providing data essential for biodiversity conservation or climate change monitoring. Similarly, observation of cetacean habitats benefits from visual surveillance to ensure ecosystem health. These applications demand high data rates for transmitting large volumes of image and video data, which optical modems can provide efficiently. Additionally, the intermediate level serves as a relay, transmitting data from deeper regions to surface nodes or sink nodes, ensuring minimal latency and high throughput.

The bottom level comprises acoustic nodes designed to transmit data over long distances but at lower bit rates. These nodes are ideal for applications requiring low-data-rate communication, such as monitoring temperature, salinity, pressure, turbidity, and/or pH levels. Acoustic communication is well suited for these applications, as it is not data-intensive and can tolerate communication delay. Acoustic nodes are deeply deployed within the observation area and transfer their data to hybrid multi-technology nodes in the intermediate level. These hybrid nodes relay the information to the top layer for further processing and dissemination. This hierarchical design ensures comprehensive coverage of underwater parameters, even in challenging environments. The integration of optical and acoustic communication systems leverages their complementary strengths, enabling high-speed data transfer in critical zones and long-range data relay for less intensive applications.

In the following sections, we will disregard the RF technology employed for the top level while discussing in detail the key protocol design issues for the intermediate and lower layers of the network architecture, which should employ UOC and UAC technologies.

4.1. Intermediate Level

The intermediate level of the underwater multi-technology architecture incorporates optical communication technologies, strategically positioned to support high-bandwidth and low-latency data transmission.

Optical nodes within this layer can be deployed at variable depths by way of depth buoys. These nodes are capable of communicating both among themselves and with an opto-acoustic gateway node, ensuring seamless integration within the overall communication system. The deployment of optical nodes significantly enhances transmission bandwidth, providing a high data rate essential for time-sensitive applications. Factors influencing the depth and area of deployment include water clarity, ambient sunlight noise, and environmental constraints [38]. For optimal performance, the positioning of optical nodes must align closely with the targeted region of interest. For example, in monitoring coral reef ecosystems or underwater archaeological sites, optical nodes are typically installed at depths corresponding to these environments. Coral reef ecosystems, often located in the photic zone, benefit from optical communication high data rate to support visual surveillance and biodiversity assessments while not implying noise generation to marine wildlife. Similarly, underwater archaeological sites may require precise and localized data collection, facilitated by optical communication nodes. The deployment depth and configuration of optical nodes are tailored to the unique requirements of each application. Key considerations include the nature of the water environment—whether shallow or deep—the technologies employed in optical communication, and operational constraints such as power supply, maintenance accessibility, and environmental impact. Additionally, water clarity plays a vital role, as optical signals are highly susceptible to attenuation due to turbidity, scattering, and light absorption. Ensuring optimal deployment strategies requires careful evaluation of these factors to balance performance and operational feasibility. Use of an intermediate optical layer calls for the design of an efficient optical network. Accordingly, in the next subsections, we detail the most critical design aspects that have to be accounted for at the various layers of the protocol stack.

4.1.1. Physical Layer Design in Underwater Optical Networks (UOCs)

The physical layer in UOCs is responsible for the reception and transmission of optical signals over the underwater medium and works as the foundation of UOCs. The major functions performed by this layer are:

• Signal Modulation and Demodulation: The Physical Layer in UOC plays a crucial role in ensuring the reliable transmission of optical signals at high data rates, even in challenging underwater environments. For effective transmission, digital data must be converted into optical signals, which can be achieved through various techniques. Among single-carrier modulation schemes, On-Off Keying (OOK), Pulse Position Modulation (PPM), and Pulse Width Modulation (PWM) are the most commonly employed, due to their simplicity, low implementation complexity, and cost-effectiveness [39,40]. However, these techniques suffer for Inter-Symbol Interference (ISI) at higher data rates. In contrast, multi-carrier modulation schemes such as Orthogonal Frequency Division Multiplexing (OFDM) mitigate ISI by transmitting data across multiple parallel sub-carriers. Both single-carrier and multi-carrier modulation schemes typically operate under Single Input Single Output (SISO) architectures, which are energy-efficient due to the simplified hardware (e.g., no need for a DAC) and lower power consumption as compared to their Multiple Input Multiple Output (MIMO) counterparts. Despite SISO simplicity, ambient light interference remains a concern; however, it can be mitigated through suppression techniques, such as optical bandpass filters or LCD-based adaptive apertures, which apply to both SISO and MIMO configurations. Meanwhile, advanced MIMO-based techniques—including Superposition Modulation (SM) and Constrained Superposition Intensity Modulation (CSIM)—offer spatial multiplexing and diversity gains, thereby boosting data rates and reliability. Nevertheless, these advantages come at the cost of increased computational complexity, synchronization requirements, and system coordination, making them less practical for power-constrained underwater networks [39,41,42].
  In addition to the previous discussion and to further investigate the impact of water quality on the optical physical layer, we present typical link parameters and performance graphs for a representative laser diode (LD)-based OOK UOC link.

  Table 4. Optical UOC link parameters and water-type coefficients (from [43]).

  Parameter	Value
  Wavelength	520 nm
  Transmit power	1 W
  Receiver bandwidth	1 MHz
  Receiver sensitivity	−53.4 dBm
  Water type (pure sea)	0.0451 m−1
  Water type (clear ocean)	0.09868 m−1
  Water type (coastal ocean)	0.31756 m−1

  outlines key parameters considered, including a wavelength of 520 nm, a transmit power of 1 W, a receiver bandwidth of 1 MHz, a sensitivity of −53.4 dBm, and extinction coefficients for pure sea, clear ocean, and coastal ocean water. Utilizing these parameters and the Beer–Lambert attenuation model [43], Figure 2 depicts the resultant received power, SNR, OOK BER, and Shannon spectral efficiency as they vary with distance for the three types of water. The graphs demonstrate that coastal ocean water causes a more rapid decline in SNR and capacity as compared to pure and clear water, and they show the typical operating ranges (tens of meters in clear water and only a few tens of meters in coastal water) for the modulation scheme under consideration.
• Channel Coding: The severe absorption and scattering in UOC lead to high attenuation, which directly impacts the BER. To address this, FEC and channel coding techniques are implemented at the physical layer of UOC. These techniques reduce BER, even in an extremely low SNR environment. FEC works by adding extra redundant bits to the transmitted signal, enabling the receiver to detect and correct errors without requiring retransmission. Commonly used FEC and channel coding techniques include Low-Density Parity-Check (LDPC) codes, Turbo Codes (TC), Polar Codes, Reed–Solomon (RS) codes, Bose–Chaudhuri–Hocquenghem (BCH) codes [39], and Trellis-Coded Modulation (TCM). Among these, LDPC and Turbo Codes offer high error correction capabilities and are particularly effective for burst error correction. However, they are computationally complex, which can be a limitation in resource-constrained underwater systems. On the other hand, Reed–Solomon (RS) and BCH codes are simpler and have lower computational costs, making them robust and easy to implement. However, their error correction capabilities are relatively lower, especially in highly challenging underwater environments with severe attenuation and scattering [44,45].
• Synchronization: In general, it is important to receive data accurately, and, accordingly, the transmitter and receiver should align in time and frequency with each other in this aim. In UOC, synchronization is particularly challenging due to the unique characteristics of the optical channel, which includes high attenuation, scattering, absorption, beam divergence, misalignment, and ambient noise. Along with these challenges, optical link’s high data rate itself demands very precise timing and frame and carrier synchronization. Different synchronization techniques such as preamble-based synchronization, pilot symbol-based, blind synchronization, frame and time synchronization sequence-based [46,47], OFDM-based (cyclic prefix and pilot subcarriers), phase locked loop, and adaptive beam steering are reported in the literature with their advantages and drawbacks. For instance, preamble-based synchronization techniques are simple and effective for burst mode data but require additional overhead and struggle against low SNR conditions. On the other hand, OFDM-based synchronization schemes perform better in multi-path scenarios and can effectively mitigate ISI by using a cyclic prefix. To enhance the efficiency of the system, advanced methods such as pilot-assisted synchronization and adaptive clock recovery methods must be preferred instead of symbol and frame synchronization alone, as they suffer more from the timing drift, jitter, and misalignment issues [48,49,50].

4.1.2. Data Link Layer Design in UOC

The data link layer in UOC should account for support of critical communication functionalities related to channel multiple access support, as well as error control, as discussed in the following.

• Multi-access protocols: UOC systems require resource allocation schemes, advanced MAC protocols, and multi-carrier transmission techniques. Various approaches have been reported in the literature, including the use of Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Code Division Multiple Access (CDMA), Wavelength Division Multiple Access (WDMA), and Space Division Multiple Access (SDMA). TDMA is among the simplest techniques, offering energy-efficient and low-latency solutions for UOC systems due to lower propagation delay [51,52]. However, TDMA-based approaches require efficient scheduling to avoid interference, whereas FDMA-based methods handle interference more effectively through non-overlapping frequency allocation. Although OFDM is not energy efficient, it has been extensively studied in UOC systems due to its robustness against ISI and superior spectral efficiency [53,54,55]. Similarly, CDMA-based schemes are promising for UOC networks due to their higher spectral efficiency, distributive access to network resources, and asynchronous operation. Extensive research highlights the potential of CDMA in UOC systems [56,57], though challenges such as power control, mobility, and coverage remain unresolved. On the other hand, WDMA reduces signal processing complexity by multiplexing multiple signals based on wavelengths [39], but it significantly increases hardware complexity and cost. While adaptive wavelength allocation could enhance efficiency, practical implementations are still lacking. In contrast to WDMA, SDMA utilizes beam directionality and spatial separation to enable simultaneous multi-user transmission, thereby improving throughput. However, it requires highly precise LoS alignment and beam steering, which is extremely challenging in underwater turbulent environments [58,59].
• Error Detection and Correction: At the Data Link Layer, error detection approaches such as CRC and the use of checksum and retransmission schemes, e.g., ARQ and HARQ, are the primary techniques for ensuring reliable frame-level communication in UOC. Methods like network coding and frame-level FEC can offer additional reliability, while adaptive techniques promise to enhance adaptability in dynamic underwater environments. These approaches complement the Physical Layer FEC schemes to provide end-to-end reliability in UOC systems.

4.1.3. Network Layer in UOC

In UOC networks, the network layer faces unique design challenges and requirements as compared to terrestrial networks. The system must ensure reliable and rapid route discovery to extend communication in a highly dynamic and directional environment, where water currents and node mobility frequently change the network topology. Additionally, node localization is essential for accurate positioning and routing. These challenges are further combined by constraints such as energy consumption, latency, and complexity, all of which are aggravated by the harsh characteristics of the underwater channel. Critical design features include:

• Connectivity: The connectivity between underwater optical nodes is inherently restrained by the dynamic environment and highly directional optical links requirement. Sparse deployment of optical nodes can further complicate the connectivity [60]. Different methods, such as optical base stations with multi-faceted transceivers [61], transmitters with a broader field of viewing angles, and tailored routing protocols, can help elevate the connectivity issues.
• Relaying and forwarding: Appropriate techniques can be implemented to extend the optical nodes’ communication range. There are several relaying methods available to overcome the inherent limitations of optical propagation underwater. Key relaying techniques include serial relaying (multi-hop transmission) and parallel relaying (cooperative transmission). Additionally, there are various packet forwarding methods such as decode-and-forward, amplify-and-forward, and bit-detect-and-forward available for use. Serial relaying uses narrow-beam optical transmitters to extend the communication range and concentrate the received signals’ power at the detector’s aperture area. However, it requires fast and accurate tracking and localization information, which might be difficult for underwater scenarios. On the other hand, parallel relaying leverages cooperation among neighboring nodes for improved diversity [62].
• Routing schemes: Underwater optical communication (UOC) networks face significant routing challenges at the network layer due to the highly directional nature of optical links and the strict angular constraints imposed by transmitter beam divergence and receiver field-of-view. Unlike acoustic systems, physical proximity between nodes does not guarantee connectivity, as valid links can only be established when precise geometric alignment conditions are met. This results in the formation of angular “dead zones,” i.e., spatial regions where nodes are physically reachable but logically disconnected from the network topology, as shown in Figure 3. Consequently, neighborhood discovery becomes asymmetric and unreliable, and conventional distance- or location-based routing strategies often fail when packets encounter routing voids caused by misaligned beams. These effects lead to fragile multi-hop paths, frequent route disruptions, and increased control overhead, making routing in UOC networks fundamentally different from and more complex than in omnidirectional underwater communication systems. The routing schemes are broadly categorized into centralized routing protocols [63,64], which rely on global network information, whereas distributed routing protocols make localized decisions to reduce the energy consumption and communication overhead [65]. Additionally, opportunistic routing [66] has emerged as a viable strategy to enhance packet delivery by exploiting the broadcast nature of optical links in dynamic underwater environments. Recently, a dual-hop routing technique [67,68] has been proposed to enhance the overall route optimization and reduce the risk of packets entering void regions as compared to single-hop [69] routing by extending topology awareness to two hops.

4.2. Lower Level

The acoustic level, positioned underwater, serves as a critical component of the multi-modal communication architecture, enabling reliable long-distance communication. Its operational range extends from a few meters to several kilometers, making it essential for applications requiring robust and consistent data transmission across vast underwater regions. Due to its capability for long-range communication at low acoustic frequencies, the acoustic level often functions as the backbone of multi-modal systems. However, the inherent limitation of low bit rates necessitates the use of efficient and lightweight networking and routing algorithms to optimize performance and ensure reliability [70,71].

This architecture in our proposal integrates two distinct types of acoustic nodes, each tailored to specific communication requirements. The first type comprises traditional acoustic nodes, which provide a set of predefined physical layer features. These nodes are designed for long-haul communication, prioritizing stability and consistency in their operational parameters. Their features ensure reliable data transmission over extended distances, making them suitable for applications where communication parameters remain relatively static.

In contrast, adding at the lower level also software-defined acoustic nodes introduces a greater degree of flexibility and adaptability. These nodes are capable of adjusting to dynamic environmental conditions and varying application requirements by offering customizable physical layer characteristics. Key parameters that can be configured include sampling rate, carrier frequency, modulation scheme, transmission power, cyclic redundancy check (CRC) size, forward error correction (FEC) methods, and node routing. This adaptability makes software-defined nodes particularly valuable in environments where conditions such as water salinity, temperature, and pressure may influence communication performance [22,33] and, thus, real-time adaptation is mandatory.

By incorporating both traditional and software-defined acoustic nodes, the acoustic level effectively balances the need for long-range, stable communication and the adaptability required for dynamic underwater environments. This dual approach enhances the overall performance and scalability of the multi-modal communication architecture, ensuring that the system can meet diverse application demands.

4.2.1. Physical Layer in Underwater Acoustic Networks

The physical layer in a UAC is responsible for the reception and transmission of signals over the underwater medium and works as the foundation of UAC networks. Following are the major challenging aspects to be taken into account in designing this layer in UACs:

• Signal Modulation and Demodulation: Since the UAC network operates in a highly dynamic environment where communication parameters frequently change, modulation schemes must be designed based on realistic scenarios that reflect the actual underwater conditions. Non-coherent modulation schemes that do not require phase synchronization information at the receiver side, such as On-Off Keying (OOK) and Frequency-Shift Keying (FSK), are examples of highly power-efficient and reliable modulation techniques, albeit achieving lower bandwidth efficiency [23]. Coherent modulation schemes are introduced to enhance bandwidth efficiency and achieve higher data rates in underwater communication networks. One example is Phase-Shift Keying (PSK), that requires the receiver to maintain synchronization with the phase of the transmitted signal. Similarly, Quadrature Amplitude Modulation (QAM), Orthogonal Frequency Division Multiplexing (OFDM), and Spread Spectrum (SS) are also classified as coherent modulation schemes [72]. Also, recently, approaches that exploit AI to perform dynamic tuning of modulation parameters based on the current channel conditions appeared. Specifically, Multi-Armed Bandit (MAB) algorithms offer a simpler yet effective solution to this selection of modulation schemes to trade off Packet Error Rate (PER) and energy consumption. This is in the view of preserving both reliability and energy efficiency [73].
• Channel Coding: The physical layer uses channel encoding techniques for error detection and correction to reduce the effects of noise, multi-path propagation, and Inter-Symbol Interference (ISI). There are different possible channel encoding schemes like convolutional codes exhibiting better performance in dynamic underwater conditions, or turbo codes that have better reliability with their parallel concatenated structures. Similarly, Low-Density Parity-Check (LDPC) codes can perform well in various underwater dynamic conditions due to their iterative decoding efficiency and flexibility in code rate and block length. However, for burst errors which are common in underwater communications, Reed–Solomon codes can perform better due to their robustness against multiple bits error handling [74].
• Synchronization: It is an essential component at the physical layer of UAC networks due to the challenges such as coping with the Doppler effect that is particularly relevant in the case of UAC systems, or multi-path propagation, and dynamic channel conditions. In an underwater dynamic environment, motion-induced fluctuations cause incoherence in timing, frequency, and phase between transmitter and receiver signals. Thus, efficient synchronization protocols must be implemented to address these issues. Synchronization is typically accomplished using techniques such as pilot tones, cyclic prefixes, and code tracking loops. These techniques, collectively, guarantee reliability and efficient data transfer in the hostile underwater environment [75].

In addition to the aspects mentioned above, other features that should be accounted for at the physical layer in UAC networks design and that can be tuned in case of software-defined modems include power control, frequency management, equalization techniques, and physical layer security mechanisms, thus ensuring optimized performance in dynamic underwater environments [76,77,78].

4.2.2. Data Link Layer in UAC

At the data link layer in underwater acoustic environments, due to the delay experienced by acoustic waves, traditional mechanisms for medium access control and error detection can be inefficient, as discussed in the following.

• Channel Access Protocols: The medium access layer is a core component in UACs for managing the channel control mechanism on a shared medium, allowing multiple nodes to communicate. Protocols such as distance aware collision avoidance protocol (DACAP) [79] and sender and receiver concurrent reservation (SRCR) [80] maximize channel utilization and improve throughput and delay, whilst coordinated transmission MAC (CT-MAC) [81] uses immediate neighbors to share information and employs a relay mechanism to obtain global information, thus optimizing energy efficiency and reducing delay time. Traditional protocols such as ALOHA and Slotted-ALOHA have problems in dealing with bursty traffic and multi-hop underwater scenarios due to the very long propagation delay. This leads to the possibility to employ handshake-based protocols such as MACA and FAMA, which use RTS/CTS signaling to reduce collisions and resolve hidden/exposed node problems, further enhanced with simultaneous RTS/CTS handshake procedures for better throughput and network efficiency [82]. However, the long propagation delay again can cause relevant inefficiency due to the need for additional signaling and the possible change in channel status as related to the carrier sense phase.
• Error Detection and Correction: The LLC layer ensures data reliability by implementing error detection mechanisms like Cyclic Redundancy Check (CRC) and error correction protocols, e.g., Automatic Repeat Request (ARQ). FEC at the physical layer requires the receiver to perform error correction and does not require any retransmission, though demanding for more computational resources. CRC, which is an error detection mechanism, adds a checksum to each data frame, allowing the receiver to recognize errors. In contrast to FEC, the data link layer employs ARQ and HARQ mechanisms to perform error correction at the transmitter side at the cost of use of too many retransmissions, which, in case of slow acoustic communications, can be unfeasible. On the other hand, use of retransmissions or FEC can compensate for the high error rate induced by noise, multi-path effects, and channel fluctuations, as typical of underwater channels [83,84].

4.2.3. Network Layer in UAC

Network layer solutions in UAC should cope with a number of issues mostly related to long propagation delay and consequent complexity in trading off reliable data delivery and topological data update. Indeed, routing in UAC is a critical task due to underlying problems like high latency, limited bandwidth, and frequent topology changes stemming from water currents and node mobility. These networks must adhere to stringent energy constraints, and dynamic propagation conditions lead to the ineffectiveness of traditional terrestrial routing protocols. Consequently, routing in this domain aims to optimize the end-to-end data delivery process through the management of energy consumption, the provision of reliable data transmission, and adaptation to the spatial distribution of nodes. Numerous routing algorithms have been designed to overcome these unique challenges—some prioritizing energy efficiency [85], others focusing on data-centric transmission techniques, and still others leveraging geographic information to optimize path selection [86]. Recently, new approaches have been developed, including reinforcement and Q-learning-based algorithms that allow adaptive adjustment of routing decisions [87,88], bio-inspired methods that leverage principles of swarm intelligence for self-organizing behaviors [85], and channel-aware solutions [89] as well as congestion-aware protocols that adjust to varying network conditions. Opportunistic routing approaches [90] further exploit the broadcast nature of underwater channels by dynamically selecting relays from a candidate set of forwarders to improve reliability and throughput. Additionally, void handling algorithms [91] address the risk of packets becoming trapped in sparse areas. Collectively, these strategies illustrate a comprehensive and innovative set of solutions to meet the multifaceted challenges of underwater acoustic communications.

5. Multi-Technology Multi-Level Testbed: A PoC at the University of Catania

In this section, we describe a PoC of a multi-level multi-technology UW network that is being developed at the University of Catania, Italy, and that has been designed according to the considerations carried out in the previous sections.

The Testbed is being deployed in the Marine Reserve Isole Ciclopi [92] in the northern part of the Catania bay, Italy (37.5614° N, 15.1664° E), where both a biological and archeological reserve exist. A map of the area is shown in Figure 4. The key environmental parameters are summarized in

Table 5. Key environmental parameters at Acitrezza Marine Reserve Isole Ciclopi.

Parameter	Value (Mean & Range)
Water Temperature	14–28 °C
Salinity	37.5–38.8 PSU
Turbidity	10–30 m
Depth Profile	20–40 m (Shelf)
pH	8.15–8.36

. The network architecture instead is reported in Figure 5.

The network works by exploiting two complementary technologies for the UW part, namely optical and acoustic communications, and considering two types of acoustic nodes, i.e., the traditional Evologics S2C18/34 modems [93] and software-defined modems developed by an Italian spinoff [94]. Outside water (top level), an RF communication relying on the use of LoRa/LoRaWAN LPWAN is used where the LoRa Dragino node [95] is located on a Bluerobotics Blueboat AUV as shown in Figure 6 [96], while the Dragino gateway is carried on top of a DJI Matrice Drone [97] so as to move the gateway closer or farther away from the LoRa node. Data format adaptation can be required at the gateway node to allow data transmission over the LPWAN sub-system. The data through the drone will eventually be forwarded over the terrestrial Internet to a remote IoT platform and/or network server and accessed through a web app or Android app, as already presented in [98,99].

At the intermediate level, a network of four LUMAX UV [30] optical modems, shown in Figure 7, is available to support transmission data rates up to 10 Mbps. Nodes are located approximatively 10 m apart from each other. Broadcast and unicast transmissions are possible at this level. The lower level consists of two coexisting networks, namely the one including 3 TD Evologics nodes (shown in Figure 8) located up to 700 m farther apart and another consisting of four SD software-defined nodes (shown in Figure 9) located at a maximum distance of 150–200 m from each other. The network of TD nodes allows the system to perform environmental parameters monitoring, thus sending only a limited amount of data compliant with the limitations to maximum rates of 13 kb/s. The parameters monitored are different, including depth, pressure (A Blue Robotics [96] Bar30 High-Resolution Depth/Pressure Sensor is used), turbidity, light intensity (a UUGear [100] Light Sensor Module is used), and pH. The control component developed for the TD nodes is in charge of controlling both the sensing part, either autonomously or according to commands received by the remote monitoring station, and the transmission equipment to properly forward the gathered data. The set of supported control functionalities on the modem includes reset of the device, power control, frame size tuning, and adaptive transmission modes. The control board, which includes a Raspberry Pi board, is also able to tune some parameters in the sensor devices, such as the compression ratio in case of image collection, the frame rate in case of recorded videos, or the sensitivity of the light sensor.

The SD nodes are employed to collect and send low-/medium-resolution pictures of the environment including fishes and ruins. The SD nodes can be programmed through a DESERT simulator script [101], thus also allowing the implementation of machine learning methodologies to select the best next relay and/or modulation scheme, as already discussed in detail in [73,102]. For example, a data-link fragmentation function is implemented which splits packets into several small frames, whose length may be varied according to the underwater channel state (short data link frames are recommended when the channel is in a bad state and the channel response varies rapidly). There are also some multi-technology nodes (inside the grey rectangles in Figure 5) where multiple pieces of equipment, consisting of sensing devices and optical and acoustic modems equipped with battery-supplied control boards, are safely connected and send data across different coexisting networks and across levels to allow contacting an edge underwater modem placed under a buoy and connected to the AUV device. The system is completely stand-alone and does not need cable connection or plugs. The coexistence among the three networks implies that some data is generated at level 2 and collected through either SD or TD nodes towards the multi-technology node on the right side. Alternatively, the presence of the other multi-technology node in the center allows the system to potentially fuse the data traveling across the SD and TD network on the left side at level 2. Then, the multi-technology node allows it to fuse and send towards the LoRa node on the AUV the flows coming from the three networks, including the optical one. This adds robustness and a degree of flexibility to the network, also to adapt to the specific application supported.

6. Challenges and Open Issues

As discussed above, designing a multi-level multi-technology architecture for underwater wireless sensor networks provides the advantages of a hierarchical network structure while also exploiting optical or acoustic technology to appropriately collect data in the most effective way, thus coping with application and environmental limitations offered by the scenario. Despite these advantages, several challenges and open research issues remain in the design of an effective methodology, as outlined in Figure 10. The latter summarizes key aspects involved in the design and deployment of multi-level, multi-technology underwater networks. These challenges arise from combining heterogeneous communication paradigms—acoustic, optical, and RF—within a hierarchical structure while operating under the severe constraints of the underwater environment. More in depth:

• Energy Harvesting and Hardware Design. The development of efficient energy harvesting techniques is crucial for the prolonged deployment of underwater nodes. Among the available methods, piezoelectric materials, bio-energy harvesting, and solar-based approaches have gained significant attention. Piezoelectric materials or strips convert underwater pressure and wave energy into electrical energy, providing a sustainable power source for underwater systems. Similarly, bio-fuel cells utilize the metabolic processes of micro-organisms to convert organic matter into electrical energy, offering an innovative and environmentally friendly solution for energy generation in underwater environments. Concerning hardware design, integration of multifaceted light detectors (receivers) can enhance link quality, reduce disruptions caused by waves and currents, and improve reception diversity.
• Deployment of Software-Defined Approaches. Software-defined optical or opto-acoustic systems would allow for dynamic adjustments of key parameters such as modulation scheme parameters, transmission power levels, coding schemes, and divergence angles in response to changes in the underwater environment. These systems could enable adaptive operation by tailoring communication settings to environmental characteristics. However, these technologies are still at an infancy phase, and only preliminary studies in this direction have been started, as in [61,103,104]. Challenges in this perspective are also posed by real-world deployments and hardware limitations.
• Routing. Routing protocols for UAC have been better studied due to the need to support communications in networks of significant extension. In contrast, UOC systems face unique challenges such as connectivity issues and link failures caused by range, alignment, and angle constraints. Also, the limited coverage range granted by the technology in underwater scenarios poses many critical requirements on the possibility to implement indeed routing among nodes. Multi-hop routing can reduce latency in UAC by leveraging the shortest path through nearby nodes, but efficient routing schemes are essential for achieving this improvement. For UOC, when designing geographic or localization-based routing combined with multi-hop strategies, it is necessary to address connectivity issues. In general, there is an urgent need for routing protocols in multi-technology architectures that incorporate void hole avoidance, fault tolerance, and congestion-aware features.
• Hardware Constraints and Optical Alignment. From a hardware perspective, underwater optical transceivers generally operate with narrow beams and receivers with a limited field of view. Consequently, even minor hardware movements, waves, and currents can easily disrupt the line of sight. Traditional gimbal control and heuristic scanning methods can increase acquisition time and energy consumption, posing significant challenges for mobile AUV-based links and long-term deployments. Thus, relaxing these stringent alignment requirements has emerged as a critical hardware issue for UOC systems.
  Recent studies have begun leveraging AI techniques to address the stringent alignment requirements of UOC systems. In [105], the alignment of optical beams between AUVs is formulated as a partially observable Markov decision process, and a soft actor–critic reinforcement learning policy is employed to maintain the link amidst navigation disturbances. In [106], a camera-based acquisition, pointing, and tracking system is implemented, which automatically re-aligns an optical link within tens of milliseconds in a bubbly underwater channel. Additionally, refs. [107,108] utilize deep neural networks and transformer-based detectors trained on dedicated underwater light-spot datasets to localize the beam and drive gimbal alignment in real time. These AI-driven alignment schemes suggest that learning-based PAT is becoming a crucial enabler for robust, high-rate optical links in dynamic underwater environments.
• Optical Channel Modeling. Although significant progress has been made in optical channel modeling, most existing studies focus on horizontal links only. Vertical communication links, especially in view of implementing multi-level architecture, must account for unique parameters, including temperature gradients, absorption, scattering, turbulence, salinity, and depth-dependent refractive index variations. The development of reliable underwater-specific turbulence channel models, in contrast to free-space optical (FSO) models, and extensive experimental validation are essential to address these challenges. Vertical UOC channel models are critical for ensuring reliable communication in underwater applications, such as deep-sea exploration and data relay between submerged nodes and surface systems [109].
• Cross-Layer Design. While substantial research exists on cross-layer designs in acoustic underwater sensor networks, opto-acoustic multi-modal systems remain underexplored. Integrating cross-layer designs in multi-technology architectures can address challenges such as limited bandwidth, high latency, and energy efficiency. By enabling inter-level information exchange, advanced protocols can be developed to optimize system performance and enhance reliability in multi-technology networks [61]. In this context, defining generic and scalable cross-layer approaches for heterogeneous acoustic–optical networks remains a key open issue. The goal is to couple the application layer QoS requirements with MAC decisions and physical-layer mode selection without introducing excessive signaling overhead or implementation complexity.

7. Conclusions

This paper presented a multi-level, multi-technology underwater network architecture that integrates optical and acoustic communication in a hierarchical design to support efficient marine monitoring. By assigning each technology to the layer where it performs best, optics for high-rate short/intermediate links and acoustics for long-range connectivity, the architecture improves scalability, reliability, and adaptability. A proof-of-concept under deployment in the Marine Reserve of Acitrezza demonstrates its practical feasibility.

Despite these benefits, several key challenges must be addressed. Energy provision remains critical, as underwater nodes rely on limited power budgets and still-maturing harvesting methods such as piezoelectric, bio-fuel, and solar strategies. So, energy-efficient design at all levels is needed. Hardware robustness is equally important: biofouling, corrosion, and motion-induced misalignment can disrupt in particular optical links, whose narrow beams require accurate pointing. On the protocol side, adaptive and software-defined operation is needed to dynamically tune parameters such as transmission power, modulation, coding, and packet length, but such intelligence demands significant computation on resource-constrained nodes. Moreover, multi-layer routing across optical and acoustic domains is complex but can be a powerful strategy in case of voids or serious channel unreliability, requiring efficient mechanisms for fault tolerance.

At the same time, the architecture offers compelling opportunities. Multi-level integration enables richer data collection across depth zones, supports high-rate applications such as video monitoring, and allows robust long-range relaying through acoustics. Hybrid nodes also enable data fusion and intelligent modality selection, paving the way for more resilient and context-aware networks. Emerging AI-driven techniques for optical alignment, routing, and cross-layer optimization promise to further enhance performance in dynamic environments.

Overall, multi-level, multi-technology underwater networks provide a strong foundation for next-generation marine monitoring systems, with clear research directions in energy efficiency, hardware reliability, adaptive protocols, and intelligent routing.