An Overview of Next-Generation Underwater Target Detection and Tracking: An Integrated Underwater Architecture

Military forces of every country are trying their best to protect their motherland from the attackers. With advancement in marine technology, it has become critical to detect and track the target by obtaining active measurements before it is close enough to attack. The utilization of unmanned underwater vehicles for target tracking behavior is gaining great attention due to continuous advancement of underwater vehicular technology. Nevertheless, safe and stable communications issues among different acoustic devices are still under active investigation to reach a robust, secure, and flexible underwater networking. Moreover, due to harsh underwater environment, acoustic simulations are also time-consuming; therefore, an accurate model for target detection and tracking is a necessity. Apart from the harsh environment of underwater networks, various technologies emerging for terrestrial networking are also becoming the part of underwater networking. For instance, cognitive acoustic networks, software-defined networks, network function virtualization, cloud computing, fog computing, and internet of underwater things; all are leading to trusted next-generation underwater networks. In this paper, we first provide a comprehensive survey of unmanned underwater vehicles and different ray tracing models essential in target detection and tracking that answers several questions regarding the current necessities of underwater networks and finally, provides a solution that opens several doors for research community to excel in this area.


I. INTRODUCTION
The networking in the ocean is an advancing technology that has been drawing abundant attention for the last two decades. A technology that allows communication among different acoustic users dealing with different applications that ranges from the depths of the ocean to the sea surface is called as underwater acoustic network (UAN). Due to the limitations of resources and complexity of ocean environment, widerange of efforts has been dedicated to tackle the problems of communications in underwater networks [1]- [4]. Similarly, the target detection and tracking in the ocean is becoming a critical issue due to increased demand of attaining manageable development of ocean resources. We consider target detection and tracking as a communication problem, since The associate editor coordinating the review of this manuscript and approving it for publication was Fang Yang. several entities need to be involved to make the detection process secure and reliable.
Target tracking is one of the important applications of underwater wireless sensor networks (UWSNs). It is a sophisticated process that estimates the state (position, velocity, acceleration) of single or multiple moving targets by conducting the possible measurements that can be available from different types of sensors. Different unmanned underwater vehicles (UUVs) can be used to detect the underwater targets that cannot be detected by fixed sensor nodes. Literature has several sonar-based target tracking algorithms that utilize underwater vehicles (mostly autonomous underwater vehicles (AUVs)) to submerge the equipment into water [5]. Sonar systems have the capability of estimation of both spatial and temporal wave fronts however, the most essential mission of sonar arrays is the estimation of direction-of-arrival (DOA), i.e. bearing information [6]. Two types of sonar systems are VOLUME 7, 2019 This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/ available for underwater target tracking: a sonar system that listen the echoes after transmitting the pulses is known as active sonar whereas the one who detects the noise made by others is called passive sonar. Active sonar systems provide both range and bearing information [7]. Three different types of active sonar systems are used to detect and track the target: monostatic (both transmitter and receiver are co-located), bistatic (receiver is separated from transmitter), and multi-static (one transmitter and multiple receivers). To estimate the target bearing and range with special diversity, a broadband signal model is used that has been proved in the simulation results of [8]. Passive sonar systems only provide bearing information. Solving bearing ambiguity is the most common issue of passive sonar systems. The systems have arrays of linear geometry that causes difficulties in distinguishing the signals; either they are originating from the right or left of the arrays [7]. Passive sonar systems are used to analyze the features of UUVs especially AUVs. Due to the advantages of target tracking based on three-dimensional (3D) measurements and sensor management, active sonar systems have received great attention over passive sonars [9], [10].
However, due to the advancement in technology, the smart next-generation underwater devices are stealthy which cause difficulty for the defense agencies to predict the sounds or echoes of the target on time. This leads to another reason that passive sonar is not suitable for robust detection in the modern next-generation underwater technology. Sound propagation in the ocean is affected by several acoustic properties of the medium. To have timely measurements, the tracer should be fast and flexible [11]. Several ray tracing models have been developed to provide acoustic simulations timely. Though, the accuracy of any ray tracing model depends on the validity and implementation of the ray theory. A few simulators/emulators (UnetStack, DESERT, WOSS, and SUNSET [12]- [15]) for underwater communication systems have been designed to conduct experiments and test the validity of different algorithms. Nonetheless, before choosing a simulator/emulator, an accurate emulation of sound propagation and channel model is a significant choice to provide an accurate picture of UAN.
Various novel techniques and standards have been proposed recently that play a crucial role in building nextgeneration underwater networks [5], [16], [17]. Likewise, terrestrial communication systems, interoperability issues due to several underwater communication approaches based on different proprietary protocols have been aroused in underwater communication systems. Underwater cognitive acoustic network (UCAN) and software-defined underwater network (SDUN) are two emerging technologies that resolve the interoperability issues in underwater communication systems. Moreover, network function virtualization (NFV) along with cloud and fog computing, the components of internet of underwater things (IoUT), are becoming popular to explore numerous underwater applications and resources and to resolve various issues (especially energy consumption and latency) in this challenging domain.
The key contributions of this paper are as follows: i A thorough review of unmanned underwater vehicles utilized for target detection and tracking and analysis of their characteristics along with existing issues are presented. ii We discuss several ray tracing models with the purpose of their development, features, and the open challenges for the designer to modify the existing issues in their codes. iii A novel next-generation underwater target detection and tracking technique is proposed to open several doors for research community to excel in this area. The proposed solution considers several acoustic devices at different control layers to integrate SDN, NFV, and fog computing in underwater networks to reduce complexities in the existing infrastructure. The remainder of the paper is organized as follows. An overview of unmanned underwater vehicles used for detecting and tracking the targets is presented in Section II. In Section III, we discuss the ray-tracing models essential in target detection and tracking, while Section IV first gives a brief introduction of each of the novel terminologies discussed in Section I, the literature using these notions to improve underwater communications, and lastly provides the solution for next-generation underwater target detection and tracking. Finally, Section V concludes the paper.

II. UNMANNED UNDERWATER VEHICLES FOR TARGET DETECTION AND TRACKING
Unmanned underwater vehicles (UUVs) are becoming part of underwater surveillance systems due to their persistency of monitoring a specific area with low costs [18]. UUVs are evolving as Navy's seaborne which is equivalent to Air Forces' drones [19]. A target tracking scheme using UUV to resolve bearing ambiguity for unique passive sonar system by executing maneuver after target detection is proposed in [7]. The system was unique because the arrays were originally designed for active sonar system and was repurposed for passive sonar system. Maneuver was executed due to the incapability of passive sonar systems to distinguish between port and starboard sides of arrays. Due to repurpose system, the performance was not good, but the system has advantage of target tracking with low cost due to not demanding any additional sensors. A coordinated scheme to track a maneuvering target by multiple UUVs with variable velocity and time delays is proposed in [20]. To achieve the goal, the information of neighboring UUV state, target state, and target acceleration is required. The scheme does not assure stability due to not considering time-varying factors of switching topology. The self-localization of UUV with a bounded estimation of its position based on signals from sources of opportunity (SOOs) is presented in [21]. The technique considers waveguide invariant and Doppler Effect 98842 VOLUME 7, 2019 and is known as waveguide invariant Doppler-based localization (WI-DBL) technique. The simulation results show that position estimation is reliable with less error in shallow water whereas deep water environment has some limitations. Three types of UUVs are equivalent to the drones of the sea: AUVs, remotely operated vehicles (ROVs), and gliders [19].

A. AUTONOMOUS UNDERWATER VEHICLES (AUVs)
AUVs seem to be a practical device for monitoring, observing and inspecting the ocean. These are autonomous vehicles that propel themselves through the water for durations of few hours to numerous days without the need of their dropping vessels. AUVs move at the speeds of almost 1.5 to 2.0 m/s on fixed trajectories. They are battery-powered vehicles with high energy density. However, high cost is required to provide enough energy for the completion of the whole mission, as duration of batteries is limited. Due to these batteries, AUVs have light-weight power sources where docking stations are required to recharge the batteries [22]. AUVs need less turning time to make 180-degree shift and may provide covert and silent operations such as mine hunting [23]. There are also some operative and deployable limitations that do not allow AUVs to move in areas of high military, shipping, or fishing activity [24]. Nevertheless, to perform several underwater tasks in a fully automated mode, AUVs are emerging as an efficient and reliable solution for various underwater complexities. The development of AUVs was started in 1990s at the research centers for gathering data and performing tasks that could not have been done in any other way [25].
To track the movement of AUVs working as receivers for multi-static sonar surveillance network, a data-driven nonmyopic approach has been proposed in [26]. The objective of the algorithm was to minimize the expected estimate error generated by the onboard tracker while locating the target. Error minimization is significant in state estimation of targets for maintaining tracks. The algorithm performs better for making decision about AUV by estimating tactical situation with a drawback of an increase in computational overhead. The results are the first successful demonstration at sea to control real-time movement of AUVs in a realistic surveillance scenario. However, a framework to deal with multiple tracks depending on the growing tactical scenarios is yet to be developed. A total of 25 experts that are part of development in adaptive mission planning (AMP) have reviewed a survey to find the answer why AUVs have not yet been deployed for AMP [27]. It was difficult for them to find clearly a single cause for failure to implement AMP; however, the main reason is the lack of demonstrations.
The authors in [28] used AUV to bridge the gap that hinders cooperation among vehicles of different architectures and modems. An AUV developed and built by the Department of Industrial Engineering of the University of Florence, MARTA AUV was particularly used to allow cooperation and data transmission among vehicles of multi-vendors modems. The experiments conducted with the help of SUNSET Software Defined Communication Stack (SDCS) framework (University of Rome La Sapienza) that opens doors for complex networking scenarios where heterogenous vehicles and systems are deployed. Another scheme proposed in [29] to solve the target tracking problem of AUVs in 3D space considered three-layer neural network and an adaptive robust controller to overcome the hindrances caused by wind, waves, and other environmental factors. The scheme assumes that AUVs are equipped with range and heading sensors, however for real implementation these sensors are of vital importance. Among several algorithms that exist in literature to track single vehicle, fleets of AUVs are also used to achieve the same goal.
A novel method [30] for the tracking of multiple AUVs is based on Probability Hypothesis Density (PHD) filter to predict the position as well as velocities of the vehicles. This is the first AUV tracker that tracks multiple vehicles in real scenarios and its validity has been approved through successful experiments. However, the scheme is limited to only underwater robotics. A passive method known as Reverse Bearing Only Target Motion Analysis (Reverse BO-TMA) for the self-localization of AUV is presented in [31] that allows AUV to maintain its distance from the source vehicle without requiring any cooperation. The accuracy of the method has been validated both numerically and experimentally; however, the method does not support inertial navigation system of AUVs. Another mathematical model is presented in [32] to analyze the probability of detection of mobile target by using AUVs. The method uses a passive sonar system to analyze the features of AUV. Likewise, for active sonar, a coherentnoncoherent joint processing framework is proposed in [8] to detect a small target in shallow water.

B. REMOTELY OPERATED VEHICLES (ROVs)
ROVs are human-controlled and tether-dependent vehicles that remain tethered to the source vessel and used mostly in applications where continuous supervision of manpower is required. These vehicles are generally equipped with cameras and can travel in the ocean as far as their tether let; therefore, they are mostly used in applications where visual information is required. They have continuous power sources used for both electrical power and communications [22]. Due to their tethered factor, ROVs draw more power and noise, and can communicate real-time data [24]. They move with less speed and have limited spatial range than AUV. They have high risks of tether's failure and are heavy weight vehicles due to the requirement of tether management system (TMS). They have high turning times typically of 4 -6 hours [33]. A 3D modelbased matching method and Real-time Multi-step Genetic Algorithm (GA) for vision-based real-time estimation of target's position and orientation by using a ROV is presented in [34]. The experimental results conducted in a pool showed that the system is robust and accurate enough to overcome environmental hindrances that not only affect the images but also the vehicular movement. An integrating scheme is proposed in [35] to emphasize the challenges of deploying, developing, and testing 3D electromagnetic sensors on different ROVs and AUVs. The objective is to enhance target detection, tracking, and classification for specific underwater defense scenarios and missions.

C. GLIDERS
Gliders have the capability to stay in the ocean for long periods with less energy consumption as compared to other two UUVs. Gliders have undulating trajectory as they propel through water using a buoyancy engine. They are not as fast as other UUVs therefore are difficult to control due to lack of power force systems. A recently developed wave glider [36] has the potential to become an important part of acoustic applications as it is mobile and could coordinate with the other ocean devices accordingly. Also, it could reduce energy costs for long durable missions. A novel hybrid heading tracking control algorithm to advance the flexibility and robustness of heading control of underwater glider is presented in [37]. Heading tracking control considers all the present and future information about the target area by reflecting the planned behavior during a mission [38]. Peter-II 200 was used to model the controller and an adaptive fuzzy incremental PID (AFIPID) along with an antiwindup (AW) compensator was considered in this algorithm. The results showed that with AFIPID and AW compensator, desired heading can be maintained even in harsh underwater environment. Due to the self-adaptation of parameters controlled by AFI system, the cost for testing underwater glider in the ocean can be reduced. A summary of all the schemes described in this section is given in Table 2.
In a nutshell, due to autonomy and the active power propelled system of AUVs, they are the best suited UUVs for underwater target tracking (see Table 1). AUVs are used to classify targets (either hidden in the depths, rocks or locate at the sea surface) due to having capability of reaching the depths of the oceans. But AUVs have the battery charging issues, which need to be considered while proposing a novel next-generation target detection and tracking scheme. Also, energy consumption is a critical issue in underwater sensor networks; therefore, integration of different novel technologies is required to solve this issue for oceanic environment. Here, we briefly discuss the existing solutions in literature for the issues of underwater energy consumption.
Ambient energy harvesting joined with supercapacitors is an up-and-coming technology to remove the necessity of using batteries and entrust only harvested energy for operational purposes [39]. Solar energy is the promising energy harvesting technique for terrestrial networks. It can be employed in IoUT with solar-powered AUVs. Solar-powered AUVs need to operate on the sea surface for battery charging via solar energy input. Other energy harvesting techniques for IoUT can be piezoelectric energy harvesting and ocean thermal energy [40]. Energy consumption is a critical issue in target tracking that can be resolved by implementing energy-efficient algorithms for target tracking. An adaptive sampling algorithm for target tracking is proposed in [41] that considers two-input-single-output fuzzy logic controller. The algorithm maximizes the energy efficiency by designing an adaptive sampling interval adjustment (ASIA) method and balances the energy consumption by developing a dynamic uncertainty threshold adjustment (DUTA) method. The authors prove 36% of energy saving in different tracking areas through simulation results.
A 3D underwater target tracking (3DUT) algorithm [9] is proposed to minimize energy consumption by incorporating a target-movement-based duty-cycle mechanism. The approach is an efficient alternate for sonar-based target tracking algorithms. Another adaptive method based on Kalman filter to minimize energy consumption in 3D underwater environment for target tracking is proposed in [42]. The method uses sleep/wake plan to track the mobile target with the trilateration method. The results show reduction in energy consumption by 33% and improvement in location error by 45%. An active detection on virtual time reversal (ADVTR) method is proposed in [43] to consider AD for channel estimation and VTR for focusing when the source-receive array (SRA) receives the reflected signal of target. Bellhop simulator was used to verify improvement in the energy of SRA and accuracy of estimating target.
All these energy-efficient protocols have been proposed with some specific designed purposes to show improvement in the overall system. None of the schemes discussed above have been considered all the aspect of target detection and tracking algorithms. The target tracking using underwater sensor networks has been gaining great attention due to low cost, medium level precision and complexity, rapid deployment, self-organized nature, real-time monitoring, and wide-range distribution [5], [44]. Therefore, we consider underwater sensor networks as a viable solution in the nextgeneration target detection and tracking integrated scheme. The integration helps resolving different issues that will be explained in Section IV.

III. RAY-TRACING MODELS
Simulators would always be chosen based on the parameters required for a user going to perform an experiment. Due to complex underwater environment, acoustics simulations are time consuming; therefore, to provide precise acoustics simulations a vital and accurate model is required. The thorough description of how sound propagates through the ocean medium is finely described by ray tracing models [45]. For short-range calculation, where the sound speed profile (SSP) and the bottom profile are considered flat, a single profile program can be used. However, for long-range calculations, where both the SSP and bottom profiles vary, an advance allpurpose program is required [46]. To determine the ray coordinates, solution of ray equations is the basic requirement of ray tracing [47]. A full 3D modeling is important for rays that travel several kilometers with vertical fluctuations. A tracer should be very fast and flexible simultaneously to handle propagations timely [11]. This section provides several ray tracing models that are available for underwater simulations. All the following models were developed to tackle the hardest problem of ray tracing i.e., the calculation of eigenrays along with other challenges [48].

A. HAMILTONIAN 3D RAY TRACING PROGRAM FOR THE OCEAN (HARPO)
HARPO is a fully 3D ray tracing program to model very long-range acoustic paths for long-term monitoring of global warming effects. This concept of long-range ocean warming was described by Munk and Forbes [49] and summarized by Gibbons [50]. The longest paths reveal some chaotic behavior that need to be suppressed by using this tracing model; however, the suppression degrades the model practicality. VOLUME 7, 2019 Also, 3D modeling is more likely to create chaotic behavior than 2D modeling; nevertheless, 3D modeling keeps the practicality of the model. HARPO has the characteristic to solve Hamilton equation on an elliptical Earth. But, these calculations of ray paths on curved Earth present complexities. Six variables are used to define ray trajectory in 3D by using non-linear Hamiltonian equations, where three define the position (radius, colatitude, longitude) and other three define the local components of the wavenumber. The interest of the developers of HARPO is to model the long-range sound propagation topology closely related to the geometry of the planet as it provides best topology. The accuracy of the model is limited to moderate ocean environment and it does not consider bathymetry effects. Also, HARPO does not generate eigenrays products (e.g., transmission loss, multipath travel time, phase, and propagation direction). Alternative software is still required to overcome the flaws of HARPO as it does not seem to be a viable model for long-range monitoring when it comes to execute it for non-smooth oceanic medium [11].

B. EIGENRAY ACOUSTIC RAY TRACING
The model was developed at Applied Physics Laboratory, University of Washington [51] and used for long-range acoustic transmission in the deep ocean. The model uses cubic spline methods to calculate SSPs and the derivative of sound speed gradient. To provide accuracy with less overhead, the model introduces prearranged user-specified step size for efficient and accurate calculations in comparison to adaptive step size. The model introduces caustics (points where intensity goes to infinity as per ray theory) when SSPs move from one profile to another in an intermittent manner, therefore a constant sound speed gradient between SSPs is an essential requirement to be implemented. The calculation of sound speed and sound speed gradient depends on 3D matrix of depth, range, and six variables at those depths and ranges. These six variables are sound speed, sound speed gradient, sound speed second derivative with depth, horizontal derivative of sound speed, horizontal derivative of sound speed gradient, and the constant. The code for this model is unstable, and it provides flexibility with an open invitation to any user for modification and accuracy. Also, a more efficient adaptive step size method and handling of SSPs at irregular depths are still required to be developed.

C. WAVE-FRONT QUEUE 3D (WaveQ3D)
WaveQ3D is a 3D ray tracing model especially designed for active sonar simulation systems and distributed as part of Under Sea Modeling Library (USML). It is based on ray theory where other models (e.g., parabolic equation and normal mode) exhibit low performance for frequencies above 1000 Hz. The objective of this model is to generate transmission loss eigenrays accurately only for coastal scenarios. Moreover, it generates other eigenrays products also i.e., multipath travel time, phase, and propagation direction. WaveQ3D enhances Gaussian beam techniques to make it applicable for lower frequencies also. These Gaussian beam techniques are based on Gaussian Ray Bundling (GRAB) [52], [23]. Due to its unique characteristic of enhances Gaussian beam techniques; it can solve Eikon equation in spherical Earth coordinates, thereby supporting out-ofplane 3D effects. Real time active sonar simulation systems exhibit better performance when the number of targets to be detected is smaller than the number of ray tracing points. However, WaveQ3D does not support this assumption for planned missions. With an increase in number of targets, WaveQ3D performs faster with less overhead based on each target. The accuracy of this model for sea surface is not yet guaranteed [54].

D. THE BELLHOP RAY TRACING MODEL
An effectual ray tracing program written in FORTRAN by Michael Porter is designed for 2D ray tracing for a given SSP and sound speed field (SSF) with different range-dependent boundaries (surface and bottom) in the ocean. Along with FORTRAN sources, MATLAB functions to present ray coordinates and transmission loss are also provided. It can be implemented in Python too [55]. The solution of dynamic ray equations is required for transmission loss or acoustic pressure. Bellhop integrates both ray and dynamic equations. The input files include depth, sound speed, surface type, attenuation, surface shape, directional sources specifications, and geo-acoustic properties. However, in the simplest case, environmental file is the only input file that includes SSP and bottom information. The output files include ray coordinates, travel time, amplitude, eigenrays, acoustic pressure, and transmission loss. The different provisions of rangedependent input and output files are shown in Table 3. Like WaveQ3D, the Gaussian beam techniques [56], [57] are used to calculate acoustic pressure in BELLHOP ray tracing model with different approximations such as, geometric beams [54], beams with ray-centered coordinates, beams with Cartesian coordinates, and Gaussian ray bundles approximation [58]. Therefore, the model can handle shadows and caustics with the help of Gaussian beam techniques. For some specific applications, where accuracy is crucial, geometric beams approximations cannot be adequate. Therefore, the model provides a set of other approximations in addition to geometric beams which is the default option. Bellhop exhibits slow performance if the number of sound speed points increases [47], [55]. A BELLHOP3D has been distributed for 3D Gaussian ray tracing model. The model has several approximations for different approaches that do not assure its accuracy nevertheless, this is an opening towards a benchmark solution for several 3D models [59].

E. HORIZONTAL -GRADIENT ACOUSTICAL RAY -TRACE PROGRAM TRIMAIN
A FORTRAN-IV program (initially written as FORTRAN-63 [60]) is used to compute intensity level, travel time, and source and receiver angles for individual eigenrays. Ray paths are calculated in a 2D medium with varying SSP. The velocity is assumed to be constant in piecewise linear functions of depth and range. The constant velocity is continuous in a triangular region however; its gradient is discontinuous at the triangular boundaries. TRIMAIN has the following features: • No limitations for SSPs are introduced in this program.
Also, SSPs are presented as a function of range and depth.
• Different rays can be traced at the same time.
• The program presents variable-bottom in piecewise linear function of depth.
• The sea surface is flat with a user-specified constant reflection coefficient and phase shift. The program uses depth, frequency, surface loss, bottom loss, and volume attenuation as input files and transmission losses, surface and bottom reflections, eigenray as output files. There are some cautions to run this program, otherwise the program ends at any point during simulations. The number of input points in SSP is limited to 50. Source depth and sound speed depth should be different. 1000 rays should be traced by the program. The program uses original ray theory. Modifications are open for any interested user. Caustics are solved in this program by throwing the eigenray out for those two rays whose depth is 0.001 meter close to each other [46].

F. VIRTUAL TIMESERIES EXPERIMENT (VirTEX)
VirTEX [61] was developed to model arbitrary motion of underwater time-varying environments and to calculate post-processing of multiple ray tracing programs (more specifically BELLHOP ray tracing program). Two novel and efficient modifications of VirTEX models to calculate single ray tracing for platform and sea surface motion respectively are: • VirTEX Extra-Lite (platform motion) • VirTEX Lite (sea surface motion) Along with post-processing of single ray tracing, VirTEX Extra-Lite accomplishes only fixed source and receiver motion modeling with less computational resources, whereas VirTEX Lite is capable of handling both steady motion of source and receiver and unsteady motion of sea surface with greater resources than VirTEX Extra-Lite. VirTEX and its variants are dependent on BELLHOP ray tracing program. The model does not consider every aspect of underwater acoustic channel.

G. THE TIME VARIABLE ACOUSTIC PROPAGATION MODEL (TV-APM)
TV-APM [62] considers realistic environment that generates channel responses for mobile nodes. The simulator which was partially developed by 7 th European Framework Program has a set of MATLAB files and data configuration files. The MATLAB files include configuration parameters that define initial positions, linear velocities, and wind parameters, whereas, data configuration files include the sea surface, bathymetry, SSP, and the transmitted signal. To simulate signal propagation dynamically, it runs Bellhop ray tracing model with the following conditions: • A 3D box with user specified bathymetry and surface wave is a must to place source and array.
• Both positions and velocities are updated with transmission time.
• A user specified SSP is required.
• A user-specified transmitted signal is required. The simulator generates the following output files: channel impulse responses and output signals. The simulator allows users to modify the parameters and configurations according to their experimental requirement. The simulator has no warranty; therefore, users are warmly welcome to fix bugs and improve the internal workings of the simulator. The simulator can handle too large velocities with the condition of enlarging run time and memory space. Shear and range dependent bottom properties, non-linear velocities and arrays, and swells are not supported by TV-APM.

H. TRACEO3D RAY TRACING MODEL
A ray tracing model developed at the Signal Processing Laboratory (SiPLAB) of the University of Algarve was designed to envisage acoustic pressure and velocity within both surface and bottom range-dependent boundaries. TRACEO was originally designed for 2D modeling and written in FORTRAN 77 however, it can surely model sound for out-of-plane propagation with simplest cases of 3D bathymetry (validity of 3D predictions). The purpose of this model is to calculate eigenrays with a specific set of SSPs or SSFs. The model can handle sound propagation in different environments with different applications; such as wavy surfaces, complex bathymetries, depth and range variations of SSPs, etc. in the areas of geo-acoustics, vector sensor arrays, and acoustic barriers, which the current existing models cannot handle. It also considers complex scenarios of earth's curvature to solve Hamiltonian equation like HARPO and Eikon equations like WaveQ3D in addition to the Lagrange formalism. It has the characteristic of solving ray equations as well as dynamic equations. The input and output files are described in Table 3. Like other models, TRACEO also handle the caustics with the help of Gaussian beam techniques. The model takes assistances from Bellhop ray tracing model, but it goes outside bellhop due to the following advancements: • Handling set of both analytical and tabular SSPs. • One or more targets can be positioned between the source and the array of receivers.
• Handling boundaries with range-dependent properties (shear velocity, attenuation) and with being absorbed by partially or totally reflective waveguide. TRACEO3D has been validated by Calazan and Rodríguez [63] as a most vital and accurate model to handle underwater noise while predicting sound propagation efficiently [45].
Numerous ray tracing models with their codes are available, but none of them satisfy all the demands of underwater acoustic monitoring for different environments. Bellhop has been considered as an accurate ray tracing model due to its characteristics of providing accurate modeling of sound 98848 VOLUME 7, 2019 propagation and acoustic channel. However, its 3D model does not assure accuracy. TRACEO3D is an advancement of Bellhop ray tracing tool as it assures out-of-plane propagation for 3D environment. Nevertheless, a fast, flexible, and accurate code is an essential requirement to handle several challenges for 3D modelling of next-generation underwater target detection and tracking architecture.

IV. NEXT-GENERATION UNDERWATER TARGET DETECTION AND TRACKING SCHEME
In this section, we first briefly discuss the novel terminologies (UCANs, SDUN, IoUT, NFV, cloud and fog computing), the literature using these notions to improve underwater communications, and then we provide the solution for nextgeneration underwater target detection and tracking.

A. UNDERWATER COGNITIVE ACOUSTIC NETWORK (UCAN)
Among various challenges of underwater environment such as severe path loss, long propagation delay, less propagation speed, limited spectrum is the one that causes hindrance in communications. Due to the high competition of sharing acoustic spectrum among various acoustic users, safe and stable communications become a great challenge for this highly challenging environment. The underwater communications systems include both natural acoustic systems (e.g., marine mammals) and artificial acoustic systems (e.g., sonar systems). The medium of communication for both natural and artificial systems is acoustic waves, thereby triggering high competition for all acoustic users to efficiently utilize the limited spectrum. Consequently, UCANs which implement cognitive acoustic (CA) capabilities are becoming popular to overcome the issues of spectrum scarcity in underwater networks. UCANs allow acoustic users (secondary users) to utilize the spectrum in a friendly manner, ensuring that primary user (licensed user) activity is secured. This means that the two acoustic users can only communicate with each other if both have consensus on common idle spectrum. But these networks are still under investigation because proposing a cognitive underwater protocol that considers the limited spectrum issue to meet the increasing demands of different acoustic users for different applications and services is highly challenging. Therefore, the research in this area is at its infancy. Very few medium access control (MAC) and network layer protocols proposed for UCANs are listed in [64]- [67].

B. SOFTWARE-DEFINED UNDERWATER NETWORK (SDUN)
Software-defined networking (SDN) has been announced as a flexible, efficient, and vendor independent technology. To engulf the constrained nature of underwater networks, the decoupling of control plane from the data plane enhances network compatibility with efficient utilization of resources and proper time management. SDN overcomes the hindrances of integration of various underwater devices to make safe and stable networking. SoftWater [68] is the first next-generation underwater communication system that integrates network function virtualization (NFV) to incorporate new underwater communication solutions. Due to this architecture, several underwater applications that operate on different communication technologies (acoustic, optical or radio waves) can be supported simultaneously. The physical-layer adaptation mechanisms are proposed in [69] to allow either seamless switching between different underwater technologies (such as orthogonal-frequency-divisionmultiplexing (OFDM) and direct-sequence-spread-spectrum (DSSS)) or joint adaptation of communication parameters (such as modulation constellation and channel coding rate). They present a new high-rate software-defined acoustic modem (SDAM) with real time adaptation capabilities and evaluate it in both indoor and outdoor environments. A multicontroller SDN-based UASN framework [70] is designed to present a load balancing mechanism with UASNs hypervisor. UASNs hypervisor is a simulation program that is developed and implemented to improve UASN performance.

C. INTERNET OF UNDERWATER THINGS (IoUT)
A network of smart-interconnected nodes that helps enabling numerous unexplored underwater applications and resources is called IoUT. The smart devices in IoUT improve numerous applications in smart coastal cities including target detection and tracking. These applications include environmental monitoring, underwater exploration, disaster prevention, military and others navigation and location applications [71]. Due to sparse network condition in IoUT, autonomous underwater vehicles (AUVs) can be used to establish stable connections between nodes far away from each other. IoUT is a threelayered architecture, first described in 2012 [40] that fulfill the demands of application users. The perception layer includes sensors, vehicles, and stations that are responsible for collecting data. The middle layer i.e., the network layer includes internet, wired/wireless networks, cloud computing platforms, and so on for transmitting information from the perception layer, and the application layer satisfies diversified demands of different users.

D. NETWORK FUNCTION VIRTUALIZATION (NFV)
NFV is a concept of virtualization of network resources into software-based network functions [17] which is at its early stage for terrestrial sensor networks. NFV provides the services in virtual machines that perform different operations. NFV can use SDN as a part of service function chaining (SFC), and SDN can provide connectivity between virtual network functions (VNFs) [72], [73]. A VNF is the virtualization of specific network function that should function autonomously. Both SDN and NFV are complementary to each other, by this means they can be applied to network of different types and can be used in simplifying network management. Therefore, for the real-time implementation of next-generation underwater sensor networks, integration of NFV and SDN facilitates multiple underwater applications under the same infrastructure.

E. CLOUD AND FOG COMPUTING
Cloud computing for underwater networks seems to be the best candidate to distribute and visualize the complex data collected from the oceans through the gateways (surface buoys or AUVs). An intelligent context-aware middleware [74] is proposed to combine underwater sensor network and cloud computing that supports different underwater applications through multi-agents. Another method [75] connects underwater sensor network to the cloud by means of a wireless transceiver to elaborate and analyze the data. An integration of ocean and cloud computing is proposed in [76] to sense, identify, and predict the multisensor data. All these research works transfer data to the cloud by means of gateways. Likewise, terrestrial communications, plethora of underwater devices increase the amount of data to be transferred to the cloud, which might cause congestion and bottleneck at the gateways. Therefore, fog computing and edge computing notions can be applied in underwater networks to overcome the issues of cloud computing. We will discuss it further in our proposed scheme below.
Taking advantages form all the novel notions such as SDN, NFV, fog, and cloud computing, we propose novel target detection and tracking scheme. The scheme considers underwater sensor nodes as the fundamental units that sense the presence of the target and report the sensing results to any of the AUV moving around the cluster of the sensor nodes. The AUV collects the sensing results, apply any information gathering algorithm such as decision-based methods or neural networks, and forward the collaborative output to the nextlevel AUV as shown in Fig. 1 to reach the local controller (any surface buoy).
Different surface buoys communicate with the SDN controller on land that is inside the fog cloud. The fog cloud serves as the local cloud to provide flexibility among different services and to preserve network latency. Each fog cloud forwards the collected data to the server cloud. This is nextgeneration hybrid underwater communications where both acoustic and radio waves serve as a medium of communications. The NFV hypervisor inside each fog cloud is responsible for launching different underwater application services such as surveillance, monitoring, safety, oceanographic, and oil inspection. SDN controller is responsible for monitoring these application services.
The integration of these different notions in nextgeneration underwater target detection and tracking systems allow any node of interest for any other service to collect information from any fog cloud and take full advantage of the whole network. This scheme allows detecting either a single or multi, mobile or fixed target, by collecting the sensing information at nearby AUVs which forwards the collected data to the local/server cloud for stronger processing. The cloud then estimates the position of the target and predicts its trajectory more accurately. The scheme does not involve a single entity; it is a hybrid communication of several underwater devices such as sensors, AUVs, surface buoys, and base stations on land, each performing the tasks in a collaborative manner. Therefore, we can say that, this is an alternative solution to detect and track the mobile target more accurately and precisely.
The problem of energy consumption for sensor nodes can be overwhelmed by utilizing the waking-up sleep mechanism [77], [78] where only those sensor nodes near the area where the target is moving, or static keep themselves activated. The other energy-efficient target detection and tracking algorithms described in Section II can be used to solve this issue. Also, SDN controller along with virtualization hypervisor, i.e. fog cloud is responsible for adjusting the energy level of each underwater sensor node [79].

V. CONCLUSION
We introduce a novel integrated solution for next-generation underwater target detection and tracking by integrating SDN, NFV, cloud and fog computing intending to resolve the existing underwater issues. AUVs are emerging as an efficient and reliable solution for various underwater complexities and will continue playing a significant role in the exploration and monitoring of underwater systems and resources. The integration allows any acoustic device of interest for any other service to collect information from any fog cloud and take full advantage of the whole network. Likewise, terrestrial networks, SDN and NFV can also be used in underwater networks to resolve energy and time constrained issues. An overview of different unmanned underwater vehicles utilized for target tracking for various algorithms in the literature is presented. We also discuss various ray-tracing models essential in target detection and tracking and provide their purpose of development along with advantages and disadvantages. It remains as our future work to test our proposed solution using TRACEO ray tracing.