- freely available
- re-usable
Sensors 2009, 9(11), 8684-8708; doi:10.3390/s91108684
Published: 29 October 2009
Abstract
: Effective solutions should be devised to handle the effects of shadow zones in Underwater Wireless Sensor Networks (UWSNs). An adaptive topology reorganization scheme that maintains connectivity in multi-hop UWSNs affected by shadow zones has been developed in the context of two Spanish-funded research projects. A mathematical model has been proposed to find the optimal location for sensors with two objectives: the minimization of the transmission loss and the maintenance of network connectivity. The theoretical analysis and the numerical evaluations reveal that our scheme reduces the transmission loss under all propagation phenomena scenarios for all water depths in UWSNs and improves the signal-to-noise ratio.1. Introduction
Underwater Wireless Communication Networks (UWCNs) are formed by sensors and Autonomous Underwater Vehicles (AUVs) interacting together to perform specific underwater applications such as collaborative monitoring or surveillance [1]. Communication quality in Underwater Wireless Sensor Networks (UWSNs) is very challenging due to the harsh characteristics of the underwater channel, such as high and variable propagation delays, limited bandwidth, high bit error rates, multipath phenomena and multipath fading [1]. In the extreme case, the spatially-variant underwater channel can cause the formation of shadow zones, which are time-variant areas where there is little signal propagation energy due to the refraction of signals by the sound speed fluctuation [2]. When the sound speed has a negative gradient just beneath the surface [3], a shadow zone is formed because the acoustic rays are refracted downward. Refraction produces shadow zones that sound waves do not penetrate because of their curvature. The sea bottom can produce a shadow zone as well, when the rays are refracted upward. Shadow zones [3] can also appear beneath the mixed layer for a source located near the ocean surface because the acoustic energy is trapped in the surface duct (see Figure 1). The shadow zone is usually bounded by the lower boundary of the surface duct and the limiting ray. Shadow zones can also appear between convergence zones. If the source is located at the same depth of the underwater sound channel axis, the shadow zone will disappear [4]. We distinguish between shallow (depth up to 100 m) and deep water. In shallow water (order of 100 meters depth) and at ranges of 3 kms shadow zones appear [2]. In deep water (order of 1,000s of meters depth) and at ranges of 10s of kms shadow zones are formed [2].
Shadow zones cause high bit error rates, losses of connectivity and dramatically impact communications performance. Some experiments show that for high frequencies signal levels are typically at least 40 dB less than those at the edges of the shadow zone [5]. For low frequencies the signal loss in the shadow zone is less severe; however, variations in received signal-to-noise ratios (SNRs) by as much as 10 dB have been observed on time scales of several hours [6]. Therefore, many authors recommend devising solutions to handle their effects [7,8].
For these reasons, in this paper we focus on communication reliability in the presence of shadow zones. We propose a distributed adaptive topology reorganization scheme that alleviates the effects of energy limitations and is able to maintain connectivity between sensor nodes in multi-hop three-dimensional UWSNs in the presence of shadow zones. Besides, it is able to estimate when the shadow zones have disappeared using double sensor units to restablish communication very quickly through the original acoustic wireless links. We study the effects of the proposed scheme in shallow (depth up to 100 m) and deep water. A two-path Rayleigh fading channel model and different propagation phenomena are considered: shallow water, deep water with convergence zones, deep water with deep sound channel and shallow or deep water with shadow zones.
According to the number of links in the three-dimensional UWSN affected by the shadow zone, three different cases have been introduced. In this paper we have extended our work in [9], where only one case was discussed. A mathematical model has been developed for each case to find the optimal placement for the sensor nodes whose ongoing communications are being disturbed. The two major objectives of the proposed mathematical model are: the minimization of the transmission loss and the maintenance of network connectivity. To the best of our knowledge this is the first paper that finds the optimal location for underwater sensor nodes affected by a shadow zone under different propagation phenomena.
The theoretical analysis and the numerical evaluations reveal that the average transmission loss values are reduced significantly under all propagation phenomena for all water depths in UWSNs when the optimal locations are computed and the communication between sensor nodes is again restablished outside the shadow zone. The average SNR values have also improved significantly and are maintained for all frequencies.
The remainder of the paper is organized as follows: In Section 2, we discuss the related work. In Section 3, we analyze our system model. In Section 4, we state the location optimization problems and propose nonlinear programming (NLP) formulations. In Section 5 we present our numerical results. Finally, we draw the conclusions in Section 6.
2. Related Work
Shadow zones represent a serious obstacle for good communication in UWSNs, because they cause network partition. Therefore, the proposal of solutions to handle their effects has been encouraged [7,8]. A distributed reactive shadow zone and delay aware (SZODAR) routing protocol for three-dimensional UWSNs was introduced in [10]. In SZODAR, the depth of the acoustic transceivers of sensor nodes is changed to avoid shadow zones, while the sensing modules are kept unchanged. During the connection establishment process, each node should send HELLO messages until it receives a routeReply or until the maximum number of allowed HELLO messages maxTrial is reached. If the number of HELLO messages is over maxTrial, the node is considered to be out of range or located in a shadow zone. In this case the node should increase its transmission power to the next power level stepwise up to the maximum and try to establish connection again at each step. Since UWSNs have very limited resources, the proposed connection establishment process consumes excessive power and can lead easily to energy depletion. In addition, it is not possible to determine if the node unable to establish connection using maximum power level is located in a shadow zone or is simply a node out of the range of the other nodes. Our proposed scheme has been designed to detect shadow zones based on the estimation of the transmission loss values, which is a not such an excessive energy consuming operation.
In addition, in [10] the acoustic transceiver of the node out of range or located in a shadow zone is moved upwards or downwards according to a random probability, whereas the node periodically tries to establish connection with its neighbours. If the boundary for the operational depth is reached and the connection establishment process has failed, the node moves in the opposite direction. With this arbitrary movement criterion this shadow zone awareness routing protocol consumes excessive power and incurs high delays and routing overhead. Moreover, the effects of different underwater propagation phenomena on the performance of the proposed protocol have not been analyzed.
In our proposed scheme sensor nodes are double units operating as a single sensor; they are decoupled into two sensor nodes in the presence of a shadow zone. One sensor node remains in the same position and estimates, as opposed to SZODAR [10], when the shadow zone has disappeared. An optimal placement imposing a restriction in energy consumption is computed for the other sensor node; this node is moved directly to this optimal location away from the shadow zone. In this way, connectivity between both sensors inside the shadow zone is established by a wire, connectivity between neighbour nodes is maintained and latency as well as power consumption are reduced. The effects of the different propagation phenomena on the performace of the proposed scheme have been analyzed. Furthermore, connection can again be reestablished very quickly using the original wireless acoustic links if the shadow zones disappear.
Several papers have proposed options related to location optimization of sensor nodes underwater. In [11], a placement scheme has been introduced to find the optimal placement of data collectors in UWSNs. In [12], a distributed node deployment technique for UWSNs to improve coverage and provide connectivity with the surface stations has been proposed and compared in terms of coverage and connectivity with the optimal centralized solutions from the literature. In [13], deployment strategies for two-dimensional and three-dimensional communication architectures for UWSNs have been developed to determine the minimum number of sensors to be deployed to achieve optimal sensing and communication coverage according to the application requirements. However, these contributions don't analyze which is the optimal location for underwater sensor nodes affected by a shadow zone under different propagation phenomena.
3. System Model Analysis
We consider a three-dimensional underwater sensor network for environmental monitoring (see Figure 2). In this architecture, a certain number of underwater sensors are deployed to cover a large area of a lake or ocean. Metal pieces at the seabed and thin ropes are used to attach the sensor nodes and distribute them at different depth levels [1]; they are deployed at different water depths to observe water quality parameters since water characteristics vary mostly in vertical direction. An electronically controlled engine at the sensors is used to change their depths [1]. This deployment strategy requires building a tree-like hierarchical multi-hop routing topology. The sensed data is sent to the surface station, which floats on the water surface and uses long distance radio communication to send this data to an onshore station. The onshore station is located at the monitoring center and performs further analysis of the data.
4. Problem Formulations and Optimization Models
We consider the example of the tree-like topology shown in Figure 2, where sensor nodes sense data and sent it upwards to the surface station using multi-hop routing. We assume that the distributed sensor nodes are double units. The cost of implementing double sensors will be limited because software modems will be coupled with generic microphones and speakers that are built-in to create off-the-shelf sensor modules [14-16]. In this way, the need for specialized communication hardware will be eliminated and the system cost will be reduced. The cost per node ($) for hardware modems that use especialized hardware is around 10^{4} $, whereas the cost of software modems that use generic hardware (off-the-shelf configuration) is around 10^{2} $; this cost decrease facilitates the deployment of sensor nodes to form underwater acoustic sensor networks. The double sensor nodes operate as a single sensor when the underwater communication is reliable. This fact changes in the presence of shadow zones.
Each intermediate node s_{i} should be able to detect if the quality of the uplink l_{i i}_{+1} towards one-hop neighbor s_{i}_{+1} is affected by the presence of a shadow zone computing the transmission loss TL_{i i}_{+1} [3]. If TL_{i i}_{+1}(t + Δt) = TL_{i i}_{+1}(t) + TL_{shadow zone} and this value persists during a period longer than T to eliminate insignificant variations, the sensor node assumes that the uplink is located in a shadow zone because TL_{i i}_{+1}(t + Δt) has been increased by the transmission loss value TL_{shadow zone} inside the shadow zone.
We distinguish between three different cases according to the number of links in the underwater sensor network affected by the shadow zone:
Case 1: Only the uplink l_{i i}_{+1} between s_{i} and s_{i}_{+1} is located in the shadow zone.
Case 2: The uplink l_{i i}_{+1} between s_{i} and s_{i}_{+1} and the uplinks l_{(}_{i-}_{1)}_{k i} between s_{(}_{i-}_{1)}_{k}, ∀k ∈ K and s_{i} are located in the shadow zone. K represents the set of neighbours connected to s_{i} through uplinks using hierarchical routing.
Case 3: The uplink l_{i i}_{+1} between s_{i} and s_{i}_{+1} and the uplinks l_{(}_{i}_{-1)}_{k i} between s_{(}_{i-}_{1)}_{k}, ∀k ∈ K and s_{i}, l_{(}_{i-}_{2)}_{k i-}_{1} between s_{(}_{i-}_{2)}_{k}, ∀k ∈ K and a particular s_{(}_{i-}_{1)}_{k}, …, l_{(}_{i-x}_{)}_{k i-x}_{+1} between s_{(}_{i-}_{x)}_{k}, ∀k ∈ K, ∀x ∈ N and a particular s_{(}_{i-x}_{+1)}_{k} are located in the shadow zone.
We assume that sensor nodes are located in a grid (see Figure 6). The axes of the coordinate system are set at the water surface. Sensors know their fixed two coordinates (x,y). They are also aware of their depth (z coordinate), which varies with time depending on the networks tasks. The z coordinate is zero at surface and increases with depth. Sensor nodes also know the location of their neighbours because they exchange their depth information.
Case 1: Only the uplink l_{i i}_{+1} between s_{i} and s_{i}_{+1} is located in the shadow zone
In this case under the presence of a shadow zone the sensor node s_{i} is uncoupled into two sensor nodes s_{i}_{→1} and s_{i}_{→2} as shown in Figure 3. These sensor nodes will be connected to each other inside the shadow zone by a wire to maintain robust communication. The node s_{i}_{→1} remains in the same location as the node s_{i} was and continues sensing data at the required depth level. We should now determine the optimal location for the node s_{i}_{→2} We formulate this problem as a Nonlinear Program (NLP).
We introduce the following notation:
u_{i}_{+1} = (x_{i}_{+1}, y_{i}_{+1}, z_{i}_{+1}) is the location of node s_{i}_{+1}.
u_{i}_{→2} = (x_{i}, y_{i}, z_{i}_{→2}) is the new location of node s_{i}_{→2}.
H is the sea depth.
R_{MAX} is the maximum transmission range.
TL_{Th} is the detection threshold.
r_{i}_{→2 i}_{+1} refers to the transmission range of the acoustic link between nodes s_{i}_{→2} and s_{i}_{+1} expressed in meters.
α represents the absorption coefficient and has the units dB/Km.
χ is a variable used to compute the transmission loss with the value χ = 10 for shallow water (cylindrical spreading) and χ = 20 for deep water (spherical spreading) [3].
A is called the transmission loss anomaly and summarizes the combined effects of several complex factors such as multiple path propagation due to the variations of the speed with temperature, depth and salinity, refraction effects, diffraction and scattering of sound by particles, bubbles and plankton within the water column. Detailed formulas of the multipath propagation loss according to the different propagation phenomena (surface reflections, bottom reflections, convergence zones, deep sound channel, etc.) can be found in [3]. It is given in dB.
w_{sz}_{−}_{start} = (x_{i}, y_{i}, z_{sz}_{−}_{start}) and w_{sz}_{−}_{end} = (x_{i}, y_{i}, z_{sz}_{−}_{end}) are the coordinates of the points delimiting the shadow zone at (x_{i}, y_{i}) with z_{sz}_{−}_{start} < z_{sz}_{−}_{end} and height z_{sz}_{−}_{end} − z_{sz}_{−}_{start}.
E_{Bw} is the energy consumed moving the sensor s_{i}_{→2}.
E_{Th} is the energy consumption threshold.
P1: Location Optimization Problem
Given: | x_{i}, x_{i}_{+1}, y_{i}, y_{i}_{+1}, z_{i}_{+1}, H, R_{MAX},TL_{TH}, α, χ, |
A, z_{sz}_{–}_{start}, z_{sz}_{–}_{end}, E_{bw}, E_{th} | |
Find: | z_{i}_{→2} ∈ [0, H], r_{i}_{→2 i}_{+1} ∈ R^{+} |
Min: | TL_{i}_{→2 i}_{+1} = χ log_{i}_{→2 i}_{+1}+α.r_{i}_{→2 i}_{+1} 10^{−3} + A |
Subject to: |
The absorption coefficient α is computed as derived in [17,18] as a function of the frequency f, the salinity S, the temperature T, the water ph pH and the depth z. We consider a two-path Rayleigh fading channel model, where each path is Rayleigh distributed. More specifically, the envelope of the signal from each path in the two-path model is modeled as an independent Rayleigh distributed random variable, α_{i}, i ∈{1,2} [19]. Consequently, for each path, the received energy per bit per noise power spectral density is given by, ψ = α^{2}.(E_{b}/N_{0}) which has a distribution, f_{Γ}(ψ)=(1/ψ_{0}). exp(-ψ/ψ_{0}) where ψ_{0} = E[α^{2}].(E_{b}/N_{0}) and E_{b}/N_{0} can be found from the SNR of the channel. Since the 2-path Rayleigh model does not have a closed-form expression for SNR distribution, it is found through simulations [19]. The value of α is used to find the transmission loss and the SNR of the channel.
Constraint (2) imposes that the transmission range should be lower than a threshold to ensure connectivity. Constraint (3) expresses that the transmission loss should be lower than a transmission loss threshold, that is, the maximum propagation loss for properly receiving the transmitted signal.
The SNR of an emitted underwater signal at the receiver can be expressed in dB by the passive sonar equation [ 3]:
The signal level SL is related to the intensity I_{t} and hence to the transmission power of the transceiver for shallow water as follows:
In deep water the equation (7) becomes:
We consider five different channel models:
Shallow water.
Deep water with convergence zones.
Deep water with deep sound channel.
Shallow water with shadow zones.
Deep water with shadow zones.
Constraint (4) imposes that the sensor node s_{i}_{→2} should be located outside the shadow zone.
Constraint (5) states that the energy consumed by the electronically controlled engine to move the sensor to the optimal position should be lower than a threshold. It is defined as [21]:
Case 2: The uplink l_{i i}_{+1} between s_{i} and s_{i}_{+1} and the uplinks l_{(}_{i-}_{1)}_{k i} between s_{(}_{i-1}_{)}_{k}, ∀k ∈ K and s_{i} are located in the shadow zone. K represents the set of neighbours connected to s_{i} through uplinks using hierarchical routing
In this case under the presence of a shadow zone the sensor node s_{i} is uncoupled into two sensor nodes s_{i}_{→1} and s_{i}_{→2}, as shown in Figure 4. These sensor nodes will be connected to each other inside the shadow zone by a wire to maintain robust communication. Now we should determine the optimal location for the nodes s_{i}_{→1} and s_{i}_{→2} using NLP. Since the topology is tree-like, we assume several nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K forward their data messages towards node s_{i}. Our objective consists of minimizing the transmission loss while preventing from network partition. For this purpose the connectivity between nodes s_{i}_{→1} and s_{(}_{i-}_{1)}_{k}, ∀k ∈ K on the one hand, and the connectivity between nodes s_{i}_{→2} and s_{i}_{+1} on the other hand should be maintained. The optimal location for the node s_{i}_{→2} is obtained solving the P1 optimization problem already introduced in Case 1. We should now obtain the optimal placement for the node s_{i}_{→1}.
We define:
u_{i}_{→1} = (x_{i}, y_{i}, z_{i}_{→1}) is the new location of node s_{i}_{→1}.
u_{(}_{i-}_{1)}_{k} = (x_{(}_{i-}_{1)}_{k}, y_{(}_{i-}_{1)}_{k}, z_{(}_{i-}_{1)}_{k}), ∀k ∈ K is the location of node s_{(}_{i-}_{1)}_{k}.
Given: | x_{i}, x_{(}_{i}_{−1)}_{k}, y_{i}, y_{(}_{i}_{−1)}_{k}, z_{(}_{i}_{−1)}_{k}, H, R_{MAX},TL_{Th}, |
α, χ, A, z_{sz}_{–}_{start}, z_{sz}_{–}_{end}, E_{bw}, E_{Th} | |
Find: | z_{i}_{→1} ∈ [0, H], r_{(}_{i}_{−1)}_{k i}_{→1} ∈ R^{+} |
Min max_{k} TL_{(}_{i}_{−1)}_{k i}_{→1} | |
Subject to: |
P2: Location Optimization Problem
The objective function of problem P2 aims at finding the best location for s_{i}_{→1} that minimizes the transmission loss of the link l_{(}_{i-}_{1)}_{k i} with the highest transmission loss value and preserves the connectivity with nodes s_{(}_{i-}_{1)}_{k} preventing from network partition.
Constraint (12) imposes that the transmission range should be lower than a threshold to ensure connectivity. Constraint (13) expresses that the transmission loss should be lower than a transmission loss threshold. Constraint (14) imposes that the sensor node s_{i}_{→1} should be located outside the shadow zone. Constraint (15) states that the energy consumed by the electronically controlled engine to move the sensor to the optimal position should be lower than a threshold.
Case 3: The uplink l_{i i}_{+1} between s_{i} and s_{i}_{+1} and the uplinks l_{(i-1)}_{k i} between s_{(}_{i-1}_{)}_{k}, ∀k ∈ K and s_{i}, l_{(}_{i}_{-2)}_{k i}_{-1} between s_{(}_{i-2}_{)}_{k}, ∀k ∈ K and a particular s_{(}_{i-}_{1)}_{k}, …, l_{(}_{i}_{-}_{x}_{)}_{k i-x}_{+1} between s_{(}_{i-x}_{)}_{k}, ∀k ∈ K, ∀x ∈ N and a particular s_{(}_{i-x}_{+1)}_{k} are located in the shadow zone
In this case the sensor node located at the lowest depth in the shadow zone s_{i} is uncoupled into two sensor nodes s_{i}_{→1} and s_{i}_{→2} as shown in Figure 5.
The optimal placement for node s_{i}_{→2} is obtained solving the P1 optimization problem already introduced in Case 1. The node s_{i}_{→1} should move to (x_{i}, y_{i}, z_{sz-end} + ε), where ε ∈ R^{+} is a very small value, that is, s_{i}_{→1} moves vertically downwards outside the shadow zone. In this case, the connectivity with the nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K through the uplinks can only be maintained if they also move outside the shadow zone. Therefore, the sensor nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K are uncoupled into two sensor nodes s_{(}_{i-}_{1)}_{k}_{→1} and s_{(}_{i}_{-1)}_{k}_{→2}. The node s_{(}_{i-}_{1)}_{k}_{→2} remains in the same location as node s_{(}_{i-}_{1)}_{k}, ∀k ∈ K was and continues sensing data at the required depth level. We should obtain the optimal placement for the node s_{(}_{i-}_{1)}_{k}_{→1}. The data sensed by s_{(}_{i-}_{1)}_{k}_{→2} is sent through a wire between s_{(}_{i-}_{1)}_{k}_{→2} and s_{(}_{i-}_{1)}_{k}_{→1} to maintain communication inside the shadow zone. Generally speaking, the same process should be repeated for all the nodes down in the hierarchy s_{(}_{i-x}_{)}_{k}, ∀k ∈ K, ∀x ∈ N located in the shadow zone. They are uncoupled into two sensor nodes s_{(}_{i}_{-}_{x}_{)}_{k}_{→1} and s_{(}_{i-x}_{)}_{k}_{→2}. The node s_{(}_{i-x}_{)}_{k}_{→2} remains in the same location as node s_{(}_{i}_{-}_{x}_{)}_{k}, ∀k ∈ K, ∀x ∈ N was and continues sensing data at the required depth level. We should obtain the optimal placement for the node s_{(}_{i-x}_{)}_{k}_{→1}. The data sensed by s_{(}_{i-x}_{)}_{k}_{→2} is sent through a wire between s_{(}_{i}_{-}_{x}_{)}_{k}_{→2} and s_{(i-x)k→1} to maintain communication inside the shadow zone.
If the node s_{(}_{i-x}_{+1)}_{k}_{→1} has already moved to a new location outside of the shadow zone and node s_{(}_{i-x}_{)}_{k}_{→1}does not know its depth z_{(}_{i-x}_{+1)}_{k}_{→1}, node s_{(}_{i-x}_{)}_{k}_{→1} can move to (x_{(}_{i-x}_{)}_{k}, y_{(}_{i-x}_{)}_{k}, z_{sz-end} + ε) and once there establish connection with node s_{(}_{i-x}_{+1)}_{k}_{→1} to find out the depth and afterwards compute its optimal location using this information.
Finally, the data sent towards the suface sink travels through the following nodes in the path using hierarchical routing: s_{(}_{i-x}_{)}_{k}_{→2}, s_{(}_{i-x}_{)}_{k}_{→1}, s_{(}_{i-x}_{+1)}_{k}_{→1},…, s_{(}_{i-}_{1)}_{k}_{→1}, s_{i}_{→1}, s_{i}_{→2}, s_{i}, s_{i}_{+1},…, and the surface sink.
Generally speaking, the optimal location of the node s_{(}_{i-x}_{)}_{k}_{→1} can be determined as follows. We define:
u_{(}_{i-x}_{)}_{k}_{→1} = (x_{(}_{i-x}_{)}_{k},y_{(}_{i-x}_{)}_{k},z_{(}_{i-x}_{)}_{k}_{→1}),∀k ∈ K, ∀x ∈ N is the new location of node s_{(}_{i-x}_{)}_{k}_{→1}.
u_{(}_{i-x}_{+1)}_{k}_{→1} = (x_{(}_{i-x}_{+1)}_{k},y_{(}_{i-x}_{+1)}_{k},z_{(}_{i-x}_{+1)}_{k}_{→1}) is the new already established location of node s_{(}_{i-x}_{+1)}_{k}_{→1} for a particular k ∈ K outside the shadow zone. We consider that node s_{(}_{i-x}_{+1)}_{k} was located originally in the shadow zone.
u_{(}_{i-x-}_{1)}_{k} = (x_{(}_{i-x}_{-1)}_{k},y_{(}_{i-x-}_{1)}_{k},z_{(}_{i-x-}_{1)}_{k}),∀k ∈ K, ∀x ∈ N is the location of node s_{(}_{i-x}_{-1)}_{k}. We consider that node s_{(}_{i-x}_{-1)}_{k} is not located in the shadow zone.
P3: Location Optimization Problem
Given: | x_{(}_{i}_{−}_{x}_{)}_{k}, x_{(}_{i}_{−}_{x}_{+1)}_{k}, x_{(}_{i}_{−}_{x}_{−1)}_{k}, y_{(}_{i}_{−}_{x}_{)}_{k}, y_{(}_{i}_{−}_{x}_{+1)}_{k}, y_{(}_{i}_{−}_{x}_{−1)}_{k} |
z_{(}_{i}_{−}_{x}_{+1)}_{k}, z_{(}_{i}_{−}_{x}_{−1)}_{k}, H, R_{MAX},TL_{Th}, α, χ, A, | |
z_{sz}_{–}_{start}, z_{sz}_{–}_{end}, E_{bw}, E_{Th} | |
Find: | z_{(}_{i}_{−}_{x}_{)}_{k}_{→1} ∈ [0, H], r_{(}_{i}_{−}_{x}_{)}_{k}_{→1 (}_{i}_{−}_{x}_{+1)}_{k}_{→1} ∈ R^{+} |
r_{(}_{i}_{−}_{x}_{−1)}_{k} _{(}_{i}_{−}_{x}_{)}_{k}_{→1} ∈ R^{+} | |
Objective 1: | |
Min: | TL_{(}_{i}_{−}_{x}_{)}_{k}_{→1 (}_{i}_{−}_{x}_{+1)}_{k}_{→1} |
Or Objective 2: | |
Min max_{k} (TL_{(}_{i}_{−}_{x}_{−1)}_{k} _{(}_{i}_{−}_{x}_{)}_{k}_{→1}, TL_{(}_{i}_{−}_{x}_{)}_{k}_{→1 (}_{i}_{−}_{x}_{+1)}_{k}_{→1}) | |
Subject to: |
5. Results
Now we study the performance of the proposed scheme under the presence of shadow zones via numerical evaluations. We distinguish between shallow and deep water. The five different channel models considered appear in Section 4. The parameters used in our evaluation are listed in Table 1. They follow the architecture of the commercial off-the-shelf underwater acoustic modem (LinkQuest) UWM3000 for shallow and UWM10000 for deep water [22].
The whole seabed is divided into a 2D square grid of equal sizes r = 4,000m for shallow water and r = 10,000 m for deep water as shown in Figure 6. The sensor nodes are located in a square grid floating at different depths. 25 sensor nodes are deployed in a 3D volume of 16,000 × 16,000 × 100 m^{3} for shallow water and of 40,000 × 40,000 × 5,000 m^{3} for deep water. In shallow water, sensor nodes are located following a hierarchical structure at the depth levels of 25, 50, 75 and 90 m. In deep water, we study the communication between sensor nodes located following a hierarchical structure at the depth levels of 300, 350, 400 and 450 m.
We consider realistic cases where shadow zones can influence the propagation of sensors. Figures 7, 8 and 9 show examples of the propagation paths for the sensor nodes in shallow water. The sound speed profile is based on the measurements conducted at the coast of Kauai, Hawaii [6], with a water depth of 100 m and a frequency of 1 kHz. The rays shown are for launch angles between −20° and 20° with a 1° increment. The corresponding ray traces in Figure 7 for a source depth of 90 m and a receiver depth of 75 m, in Figure 8 for a source depth of 75 m and a receiver depth of 50 m and in Figure 9 for a source depth of 50 m and a receiver depth of 25 m are plotted. Shadow zones appear at ranges around 3 Kms. In our simulations we consider a 2D square grid of equal sizes r = 4,000 m because at these ranges it is possible to appreciate the shadow zone effects.
Figures 10, 11 and 12 show examples of the propagation paths for the sensor nodes in deep water. We consider a frequency of 1 kHz and a deep-water (5,000 m) environment with a Munk sound speed profile [2]. The rays shown are for launch angles between −20° and 20° with a 1° increment. The corresponding ray trace in Figure 10 for a source depth of 450 m and a receiver depth of 400 m, in Figure 11 for a source depth of 400 m and a receiver depth of 350 m and in Figure 12 for a source depth of 350 m and a receiver depth of 300 m are plotted. Shadow zones appear at ranges around 10 Kms. In our simulations we consider a 2D square grid of equal sizes r = 10,000 m because at these ranges it is possible to appreciate the shadow zone effects.
Transmisson Loss Threshold
Now we compute the transmission loss threshold TL_{Th} or maximum propagation loss for properly receiving the transmitted signal given by (9). The link quality in underwater communication is severely affected by multipath phenomena, multipath fading and the refractive properties of the sound channel [3]. As a result, the bit error rates of the acoustic links are often high and efficient error control schemes are needed. The error control schemes automatic repeat request (ARQ), forward error correction (FEC) block codes and FEC convolutional codes have been selected and evaluated in terms of the packet error rate (PER).
According to [3], OFDM transmission with QAM modulation has been used for our analysis. The scalable OFDM with 16-QAM modulation has been successfully tested in the settings with bandwidths 12 kHz, 25 kHz and 50 kHz, leading to data rates about 12 kbps, 25 kbps and 50 kbps respectively [23]. The BER for the modulation scheme 16-QAM is given by:
Based on the bit error rate p_{b}, the PER for the error control schemes can be calculated as follows:
For ARQ, the cyclic redundandy check (CRC) block code detection mechanism is deployed. Assuming detection of all possible packet errors, the PER of a single transmission for a packet of l bits is computed as:
For FEC convolutional codes, the PER of a single transmission for a packet of l bits is given by:
The relationship between the PER and TL_{Th} for ARQ, three BCH, three RS and two convolutional codes is shown in Figure 13 for shallow water and in Figure 14 for deep water. For a target PER, FEC block codes can support TL_{Th} values higher than ARQ and convolutional codes. We notice that the maximum TL_{Th} value is increased with the error correcting capability of the FEC block codes and the code rate of the FEC convolutional codes. For a particular TL_{Th} value, the PER is equal or lower for error control codes in deep water in comparison with shallow water.
The FEC convolutional code with R_{c} = 1/2 has the lowest TL_{Th} value. This is the worst case, because the maximum transmission loss is restricted to lower values. Therefore, with this error control scheme for a target PER of 10^{−2} the maximum TL_{Th} of 112.5 dB for shallow water and 129 dB for deep water have been selected. More details about the performance of error control schemes in UWSNs can be found in [20].
Energy Consumption in the Sensor Movement
Now we analyze the energy consumed by the electronically controlled engine to move the sensors affected by the shadow zone to the optimal position. Figure 15 shows the energy consumption in the sensor movement. We observe that the energy consumed is increased with the movement power and the covered distance to reach the optimal location. Since the sensor relocation can be costly in terms of energy consumption, it is limited by the threshold E_{Th}; this paramenter is estimated analyzing the battery capacity, the required average displacement as well as P_{Bw}.
Case 1: Optimal Placement Evaluation
We consider a sensor node s_{i} located at (x_{i}, y_{i}, z_{i}). Node s_{i}_{+1} is located at (x_{i} ± j.r, y_{i} ± j.r, z_{i}_{+1}), j ∈ N, r = 4,000 m for shallow water, r = 10,000 m for deep water, z_{i}_{+1} < z_{i}. The acoustic link l_{i i}_{+1} is affected by a shadow zone with height (z_{sz}_{−}_{end} − z_{sz}_{−}_{start}) = 9 m, where z_{i} > z_{sz}_{−}_{end} and z_{i}_{+1} < z_{sz}_{−}_{start}. Node s_{i} is uncoupled into sensor nodes s_{i}_{→1} and s_{i}_{→2} and the optimal location of node s_{i}_{→2} has been found using NLP solving the optimization problem P1.
Figure 16 shows the average transmission loss as a function of frequency for shallow and deep water under different propagation phenomena. We have computed the transmission loss for the acoustic link l_{i i}_{+1} (between s_{i} and s_{i}_{+1}) affected by the shadow zone (shallow water + shadow zone or deep water + shadow zone). s_{i} is uncoupled and the optimal location for node s_{i}_{→2} is found outside the shadow zone using our mathematical model; the transmission loss for the acoustic link l_{i}_{→2 i}_{+1} (between s_{i}_{→2} and s_{i}_{+1}) (shallow water (optimal), deep water + convergence zone (optimal) or deep water + deep sound channel (optimal) has also been computed. The average transmission loss is increased when the frequency is increased and is higher for deep water + shadow zone than for shallow water + shadow zone. It is also higher for deep than shallow water. The transmission loss values are higher than the transmission loss threshold for deep water+shadow zone, which means that the transmitted signal will not be properly received. The average transmission loss values are reduced significantly when the optimal locations are computed and the communication between sensor nodes is again restablished outside the shadow zone. The maximum transmission loss improvement is of 4.9 dB for shallow water, of 45.1 dB for deep water + convergence zone and of 42.9 dB for deep water + deep sound channel. Therefore, we can conclude that using our mathematical model deep water + convergence zone shows the best improvement in the diminishment of the transmission loss.
Case 2: Optimal Placement Evaluation
We consider a sensor node s_{i} located at (x_{i}, y_{i}, z_{i}). Node s_{i}_{+1} is located at (x_{i} ± j.r, y_{i} ± j.r, z_{i}_{+1}), j ∈ N, r = 4,000 m for shallow water, r = 10,000 m for deep water, z_{i}_{+1} < z_{i}. Nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K are located at (x_{i}±j.r, y_{i}±j.r, z_{(}_{i}_{-1)}_{k}), j ∈ N, r = 4,000 m for shallow water, r = 10,000 m for deep water, z_{(}_{i}_{-1)}_{k} > z_{i}, k ∈ K. The acoustic links l_{i i}_{+1} and l_{(}_{i-}_{1)}_{k i} are affected by a shadow zone with height (z_{sz}_{−}_{end} − z_{sz}_{−}_{start}), where z_{sz}_{−}_{start} < z_{i} < z_{sz}_{−}_{end}, z_{(}_{i}_{-1)}_{k} > z_{sz}_{−}_{end}, z_{i}_{+1} < z_{sz}_{−}_{start}. Node s_{i} is uncoupled into sensor nodes s_{i}_{→1} and s_{i}_{→2}. The optimal locations of node s_{i}_{→2} and s_{i}_{→1} have been found using NLP solving the optimization problems P1 and P2, respectively. Now we evaluate the performance of the proposed scheme for the optimal location of node s_{i}_{→1} (Case 2). For shallow water the value of z_{sz}_{−}_{end} varies between 50 and 85 meters, z_{sz}_{−}_{start} = 40 m and z_{(}_{i}_{-1)}_{k} ∈ (z_{sz}_{−}_{end},100]. The average transmission loss of the acoustic link l_{(}_{i-}_{1)}_{k i} as a function of the shadow zone height (z_{sz}_{−}_{end} − z_{sz}_{−}_{start}) for shallow water and three different frequencies has been computed. The transmission loss of the link l_{(}_{i-}_{1)}_{k i} between the node s_{(}_{i-}_{1)}_{k}, ∀k ∈ K and s_{i}_{→1} with the highest value is minimized for shallow water. The transmission loss values are not affected by the shadow zone height but they are increased with the frequency to 41 dB for f =0.1 kHz, 41.6 dB for f =1 kHz and 49.3 dB for f =10 kHz. Figure 17 shows the average transmission loss of the acoustic link l_{(}_{i-}_{1)}_{k i} as a function of the shadow zone height (z_{sz}_{−}_{end} − z_{sz}_{−}_{start}) for deep water and three different frequencies. For deep water the value of z_{sz}_{−}_{end} varies between 500 and 4000 meters, z_{sz}_{−}_{start} = 440m and z_{(}_{i}_{-1)}_{k} ∈ (z_{sz}_{−}_{end},5,000]. The transmission loss of the link l_{(}_{i-}_{1)}_{k i} between the node s_{(}_{i-}_{1)}_{k}, ∀k ∈ K and s_{i}_{→1} with the highest value is minimized for different propagation phenomena (deep water + deep sound channel, deep water + convergence zone). Deep water + convergence zone is the underwater propagation phenomena that suffers lower transmission loss and is more appropriate for underwater communication.
The transmission loss values are especially decreased with the shadow zone height for the frequency of 10 KHz; the reason is that the depth of the optimal location is increased when the shadow zone height is increased (z_{sz}_{−}_{end} is higher) and the transmission loss decreases when the depth is increased. Otherwise, the transmission loss values for 0.1 KHz and 1 KHz are only very slightly decreased by the shadow zone height and are only significantly increased with the frequency.
Case 3: Optimal Placement Evaluation
We consider a sensor node s_{i} located at (x_{i}, y_{i}, z_{i}). Nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K are located at (x_{i} ± j.r, y_{i} ± j.r, z_{(}_{i}_{-1)}_{k}), j ∈ N, r = 4,000 m for shallow water, r = 10,000 m for deep water, z_{(}_{i}_{-1)}_{k} > z_{i}, k ∈ K. Nodes s_{(}_{i-}_{2)}_{k}, ∀k ∈ K are located at (x_{i} ± j.r, y_{i} ± j.r, z_{(}_{i}_{-2)}_{k}), j ∈ N, r = 4,000 m for shallow water, r = 10,000 m for deep water, z_{(}_{i}_{-2)}_{k} > z_{(}_{i}_{-1)}_{k}, k ∈ K. Nodes s_{(}_{i-}_{3)}_{k}, ∀k ∈ K are located at (x_{i} ± j.r, y_{i} ± j.r, z_{(}_{i}_{-3)}_{k}), j ∈ N, r = 4,000 m for shallow water, r = 10,000 m for deep water, z_{(}_{i}_{-3)}_{k} > z_{(}_{i}_{-2)}_{k}, k ∈ K. The acoustic links l_{(}_{i-}_{1)}_{k i}, l_{(}_{i-}_{2)}_{k} _{(}_{i-}_{1)}_{k}, l_{(}_{i-}_{3)}_{k} _{(}_{i-}_{2)}_{k} are affected by a shadow zone with height (z_{sz}_{−}_{end} − z_{sz}_{−}_{start}), where z_{sz}_{−}_{start} < z_{i} < z_{sz}_{−}_{end}, z_{sz}_{−}_{start} < z_{(}_{i}_{-1)}_{k} < z_{sz}_{−}_{end}, z_{sz}_{−}_{start} < z_{(}_{i}_{-2)}_{k} < z_{sz}_{−}_{end}, z_{(}_{i}_{-3)}_{k} > z_{sz}_{−}_{end}. Node s_{i} is uncoupled into sensor nodes s_{i}_{→1} and s_{i}_{→2}. Nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K are uncoupled into sensor nodes s_{(}_{i-}_{1)}_{k}_{→1} and s_{(}_{i-}_{1)}_{k}_{→2}. Nodes s_{(}_{i-}_{2)}_{k}, ∀k ∈ K are uncoupled into sensor nodes s_{(}_{i-}_{2)}_{k}_{→1} and s_{(}_{i-}_{2)}_{k}_{→2}. Node s_{i}_{→1} is relocated to (x_{i}, y_{i}, z_{sz-end} +ε), where ε ∈ R^{+}is a very small value, z_{sz}_{−}_{end} = 89 m for shallow water and z_{sz}_{−}_{end} = 4049 m for deep water. The optimal locations of nodes s_{(}_{i-}_{1)}_{k}_{→1} have been found with NLP solving the optimization problem P3 with the objective function 1. The optimal locations of nodes s_{(}_{i-}_{2)}_{k}_{→1} have been found with NLP solving the optimization problem P3 with the objective function 2. In shallow water, sensor nodes are located following a hierarchical structure at the depth levels of 25, 50, 75 and 90 meters. In deep water, sensor nodes are located following a hierarchical structure at the depth levels of 3,900, 3,950, 4,000 and 4,050 meters.
We consider four hierarchical levels with one node s_{i} at the lowest level, three nodes s_{(}_{i-}_{1)}_{k}, ∀k ∈ K at the next level, six nodes s_{(}_{i-}_{2)}_{k}, ∀k ∈ K at the next level and 12 nodes s_{(}_{i-}_{3)}_{k}, ∀k ∈ K at the deepest level. Figure 18 shows the average SNR for the acoustic links l_{(}_{i-}_{1)}_{k i}, l_{(}_{i-}_{2)}_{k} _{(}_{i-}_{1)}_{k}, l_{(}_{i-}_{3)}_{k} _{(}_{i-}_{2)}_{k} affected by a shadow zone as a function of frequency for shallow and deep water under different propagation phenomena. The average SNR values in the presence of shadow zones are negative, which indicates that the signal is below the noise level and transmission loss, and it can't be properly recovered. They have not been depicted but are decreased when the frequency is increased and are lower for deep water + shadow zone than for shallow water + shadow zone because the transmission loss values are higher for deep water + shadow zone. The average SNR values are increased significantly when the optimal locations are computed and the communication between sensor nodes is again restablished outside the shadow zone. Furthermore, the improved SNR values are maintained for all frequencies. With shallow water the SNR values are better because in shallow water the transmission loss values are lower. The SNR values are higher with deep water + convergence zone than with deep water + deep sound channel. Nevertheless, in all cases the SNR values are positive and they are increased outside the shadow zones. The maximum SNR improvement is very significant: 197.64 dB for shallow water, 352.08 dB for deep water + convergence zone and 350.68 dB for deep water + deep sound channel.
6. Conclusions
In this paper, a distributed adaptive topology reorganization scheme that maintains connectivity in multi-hop UWSNs affected by shadow zones has been developed in the context of two Spanish-funded research projects. It alleviates the effect of energy limitations and is able to maintain network connectivity in multi-hop three-dimensional UWSNs in the presence of shadow zones solving a location optimization problem.
Three different cases have been determined according to the number of links in the three-dimensional UWSN affected by the shadow zone. A mathematical model has been developed for each case to find the optimal placement for the sensor nodes, whose ongoing communications are being disturbed.
The average transmission loss values are reduced significantly (especially for deep water + convergence zone) when the optimal locations are computed and the communication between sensor nodes is again reestablished outside the shadow zone. Shallow water shows lower transmission loss values in comparison with the other propagation phenomena. Deep water + convergence zone suffers lower transmission loss values than deep water + deep sound channel and is more appropriate for underwater communication. Only in deep water the transmission loss values are especially decreased with the shadow zone height for the frequency of 10 KHz and only very slightly decreased for the frequencies of 0.1 KHz and 1 KHz; the transmission loss values don't vary with the shadow zone height in shallow water.
The average SNR values in the presence of shadow zones are negative, which indicates that the signal is below the noise level and transmission loss, and it can't be properly recovered. The average SNR values are increased significantly when the optimal locations are computed and the communication between underwater sensors is again reestablished outside the shadow zone. Furthermore, the improved SNR values are maintained for all frequencies. With shallow water the SNR values are better because in shallow water the transmission loss values are lower. The SNR values are higher with deep water + convergence zone than with deep water + deep sound channel. Nevertheless, in all cases the SNR loss values are increased outside the shadow zones.
This work was supported by the Spanish Government through projects TSI2006-13380-C02-01 and TSI2007-66637-C02-01.
References and Notes
- Akyildiz, I.F.; Pompili, D.; Melodia, T. Underwater Acoustic Sensor Networks: Research Challenges. Ad Hoc Networks 2005, 3, 257–279. [Google Scholar]
- Preisig, J. Acoustic Propagation Considerations for Underwater Acoustic Communications Network Development. ACM SIGMOBILE Mob. Comput. Commun. Rev. 2007, 11, 2–10. [Google Scholar]
- Domingo, M.C. Overview of Channel Models for Underwater Wireless Communication Networks. Phys. Commun. 2008, 1, 163–182. [Google Scholar]
- Brekhovskikh, L.M.; Lysanov, Y.P. Fundamentals of Acoustics, 3rd ed.; Springer: New York, USA, 2003. [Google Scholar]
- Kinsler, L.E.; Frey, A.R.; Coppens, A.B.; Sanders, J.V. Fundamentals of Acoustics, 4th ed.; John Wiley & Sons: New York, NY, USA, 2000. [Google Scholar]
- Siderius, M.; Porter, M.; the KauaiEx Group. Impact of Thermocline Variability on Underwater Acoustic Communications: Results from KauaiEx. Proceedings of High Frequency Ocean Acoustics: High Frequency Ocean Acoustics Conference, American Institute of Physics, La Jolla, CA, USA, March 1-5, 2004; pp. 358–365.
- Akyildiz, I.F.; Pompili, D.; Melodia, T. State of the Art in Protocol Research for Underwater Acoustic Sensor Networks. ACM SIGMOBILE Mob. Comput. Commun. Rev. 2007, 11, 11–22. [Google Scholar]
- Shi, H.; Kruger, D.; Nickerson, J.V. Incorporating Environmental Information into Underwater Acoustic Sensor Coverage Estimation in Estuaries. Proceedings of 2007 AFCEA/IEEE Military Communication Conference (MILCOM), Orlando, FL, USA, October 29-31, 2007.
- Domingo, M.C. Optimal Placement of Wireless Nodes in Underwater Wireless Sensor Networks with Shadow Zones. In IFIP Wirel Days; Paris, France, 15-17 December 2009. in press. [Google Scholar]
- Nguyen, S.T.; Cayirci, E.; Yang, L.; Rong, C. A Shadow Zone Aware Routing Protocol for Acoustic Underwater Sensor Networks. IEEE Commun. Lett. 2009, 13, 366–368. [Google Scholar]
- Alsalih, W; Akl, S; Hassanein, H. Placement of Multiple Mobile Data Collectors in Underwater Acoustic Sensor Networks. Proceedings of IEEE International Conference on Communications (ICC), Beijing, China, May 19-23, 2008.
- Akkaya, K.; Newell, A. Self-Deployment of Sensors for Maximized Coverage in Underwater Acoustic Sensor Networks. Comput. Commun. 2009, 32, 1233–1244. [Google Scholar]
- Pompili, P; Melodia, T; Akyildiz, I. Three-Dimensional and Two-Dimensional Deployment Analysis of Underwater Acoustic Sensor Networks. Ad Hoc Networks 2009, 4, 778–790. [Google Scholar]
- Jurdak, R.; Lopes, C.V.; Baldi, P. Software Acoustic Modems for Short Range Mote-Based Underwater Sensor Networks. Proceedings of IEEE Oceans, Swissotel, The Stamford, Singapore, May 16-19, 2006.
- Jurdak, R.; Aguiar, P.M.Q.; Baldi, P.; Lopes, C.V. Software Modems for Underwater Sensor Networks. Proceedings of IEEE/OES Oceans'07, Aberdeen, Scotland, June 18-21, 2007.
- Jurdak, R.; Ruzzelli, A.G.; O'Hare, G.M.P.; Lopes, C.V. Mote-Based Underwater Sensor Networks: Opportunities, Challenges, and Guidelines. Telecommun. Syst. (Kluwer Publishers) 2008, 37, 37–47. [Google Scholar]
- Francois, R.E.; Garrison, G.R. Sound Absorption Based on Ocean Measurements: Part I: Pure Water and Magnesium Sulfate Contributions. J. Acoust. Soc. Amer. 1982, 72, 896–907. [Google Scholar]
- Francois, R.E.; Garrison, G. R. Sound Absorption Based on Ocean Measurements: Part II: Boric Acid Contribution and Equation for Total Absorption. J. Acoust. Soc. Amer. 1982, 72, 1879–1890. [Google Scholar]
- Vuran, M.C.; Akyildiz, I.F. Cross-layer Packet Size Optimization for Wireless Terrestrial, Underwater, and Underground Sensor Networks. Proceedings of IEEE INFOCOM '08, Phoenix, AZ, USA, April 13-18, 2008.
- Domingo, M.C.; Vuran, M.C. Cross-Layer Analysis of Error Control in Underwater Wireless Sensor Networks. Unpublished work. 2009. [Google Scholar]
- Wildi, T. Electrical Machines Drives and Power Systems, 6th ed.; Prentice Hall: Upper Saddle River, NJ, USA, 2005. [Google Scholar]
- LinkQuest. Underwater Acoustic Modems, Available online: http://www.link-quest.com (accessed August, 2009).
- Li, B.; Zhou, S.; Huang, J.; Willett, P. Scalable OFDM Design for Underwater Acoustic Communications. Proceedings of International Conference on Acoustics, Speech and Signal Processing (ICASSP), Las Vegas, NV, USA, March 30–April 4, 2008.
Table 1. Parameter Values. |
Parameter | Value |
---|---|
Sea depth H | Shallow water: 100 m |
Deep water: 5,000 m | |
Volume | Shallow water: 16,000 × 16,000 × 100 m^{3} |
Deep water: | |
40,000 × 40,000 × 5,000 m^{3} | |
Number of sensors | 25 |
Packet size | 238 bytes |
Data rate | 6 Kbps |
T | 15 °C |
pH | 8 |
S | 35 ppt |
Transmission power P_{t} | Shallow water: 12 W |
Deep water: 40 W | |
Noise level | 47.69 dB |
Speed of sound | 1500 m/s |
Sensor speed | 0.5 m/s |
Movement power P_{Bw} | 10 W |
Λ | 0.01 |
R_{MAX} | Shallow water: 5,000 m |
Deep water: 11,000 m | |
E_{Th} | 1,500 J |
TL_{Th} | Shallow water: 112.5 dB |
Deep water: 129 dB |
© 2009 by the authors; licensee Molecular Diversity Preservation International, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).