In this section, the detailed simulation analysis of the proposed scheme in contrast to the conventional AHH-VBR scheme is presented. To fairly evaluate the performance of both the schemes, we simulated an underwater 3D network of 2 km × 2 km × 4 km area, where 2 km and depth of 4 km. The network size of 200 to 450 nodes has been simulated with the varying transmission ranges, ranging from 500 m to 900 m to demonstrate the sparse and dense network scenarios. In each simulation trial, the nodes are deployed randomly in the said network area and every individual sensor node acts as a data source (generates data packets) as well as a forwarder node. The position of the Sink node is static during the whole simulation course. Sink node is positioned at the water surface and at the center of the network area with coordinates (, , 0). All nodes are homogeneous in terms of transmission range in every trial and are assigned initial energy as , where and . A single network scenario for a given transmission range and network size is simulated 100 times. Therefore, all distinct points in the graphs of the simulation results are an average of 100 simulation trials.
The payload size of the data packet, neighbor request, and acknowledgment packets are bits, 64 bits, and 112 bits, respectively. The common header of 88 bits is used for all packet types in simulation. In addition to that, the data rate of bits per second and the underwater acoustic delay propagation delay of 1500 m has been set in the simulations. Network is static during the complete simulation period. In last, the pure ALOHA is used at MAC layer because it is not susceptible to delays and does not use any collision detection and the avoidance mechanism.
As stated earlier in a brief discussion about the conventional AHH-VBF, it ensures the packet forwarding reliability by setting the minimum forwarder threshold, , which depends upon the error probability, the packet collision rate, and the size of the packet. However, in simulations, AHH-VBF considered , which indicates that there should be at least two or more than two forwarders in the forwarding region. This ensures the reliability as well as the collision probability at the next hop forwarder(s). Additionally, it consumes more energy and utilizes more bandwidth, which is the scarce resource of the UASN. This situation can easily arise when the holding time difference between two or more than two forwarders is very negligible or smaller than the propagation delay between them. On the contrary, our proposed scheme intends to avoid multiple transmissions of the data message towards the upstream direction, to save energy and avoid collision at upstream receivers. In short, the proposed scheme not only selects the spatially suitable but also the energy-rich acoustic node among the forwarders pool. Hence, the fair performance comparison is achieved by setting and analyzing the packet suppression count or the number of forwarders count and energy consumption. During all simulation scenarios, there are 200 data sources that generate data packets destined to Sink node. The same number of data sources is used for the large network size scenarios with the intention to find the impact of identical data traffic on network performance.
4.2. Simulation Results in the Static Sink Scenario
In this section, all the results are estimated for the network scenario with static sink. In this case, the Sink is placed at sea surface and the center of the network deployment region.
and Figure 6
show the total number of copies of the data message forwarded in the network versus verying network size and transmission range, respectively. It can be seen in Figure 5
that in a sparse network scenario,
and network size, the total copies of data packet forwarded in the network is smaller than the dense network scenario. The reason behind this phenomenon is that most of the copies of the data packet fail to reach the next hop forwarders. As the network size increases, more copies of the data message are successfully propagated in the network. On the contrary, for large transmission range and small network size, more data packets are successfully forwarded in the network, but when the network size increases, most of the packets are dropped in the network due to a large collision probability. Increase in data packet copies due to large transmission range plus network size is due to two factors; (a) increase in the pipeline radius directly surges in the number of forwarding candidates in PFZ, (b) the propagation delay between potential forwarders in PFZ will be longer and could be more than the holding time difference between them.
The next interesting trend that has been observed in the graphs is the difference between total copies of the data packet that are forwarded by AHH-VBF is larger than the proposed scheme. It shows that the holding time difference between the potential forwarders is less than the propagation delay between those nodes. Hence, multiple nodes from the PFZ send copy of the same data packet, which results in more energy consumption that is discussed later in this section. In contrast, the proposed scheme computes holding time and exponentially scales the holding time using normalized energy factor to increase the holding time difference between potential forwarders. Consequently, it minimizes the data packet duplication and saves energy in the proposed scheme. The same performance metric has been investigated for different transmission ranges and fixed network size and a similar trend has been observed in Figure 6
. On average, the proposed scheme generates
less copies of the data packet for all
and network size of 450 nodes. Similarly, in
m and all network size scenarios, the proposed scheme disseminates about
fewer copies of the data packet in the network. A detailed performance gain achieved by the proposed scheme in this regard is shown in Figure 7
. The maximum performance gain achieved by the ESEVBF in contrast to the AHH-VBF is
for the network size of 450 nodes and
Next, we investigate the end-to-end delay experienced by the successful data packets between the source S
and sink node D
for varying network size and transmission range as shown in Figure 8
and Figure 9
, respectively. The overall delay experienced by a data packet includes processing, propagation, and the holding time delay at each forwarding stage in the network. The impact of network sparseness and denseness can be seen in those graphs. It is evident from the graphs that the end-to-end delay experienced by the AHH-VBF is larger than the proposed scheme. The main reason behind this behavior is our proposed potential forwarder selection algorithm. As stated earlier in the proposed forwarder selection algorithm section, all the nodes share their location and residual energy information with their 1-hop neighbors. When a node receives data packet from any one of its neighbors, it computes its own holding time. As we recall from our previous discussion that the holding time computation only requires the location plus depth information of sender node S
, Sink node D
, node’s own location, and the residual energy information. However, the same information of the neighboring nodes is also available in the neighborhood table of each node. Hence, the node can easily estimate the packet holding time of all common neighbors of S
and node itself, which also fall in the PFZ
It is also a well-established fact that computation energy cost is very small compared to communication and other operations of the node. Once the node estimates the holding time of its own as well as its neighbors, it checks whether its own holding time is smaller than the neighbors or not. If the receiving node’s holding time is smaller than its common neighbors, then instead of waiting for a long holding time duration, it forwards the packet after
. However, the AHH-VBF does not exploit the neighborhood information available at the node and each node has to wait for a holding time duration before further relaying the data packet. This is the reason that despite the energy scaling and expansion of the proposed holding time, its end-to-end delay is smaller than the AHH-VBF. Subsequently, the same behavior can be observed for any network scenario, refer Figure 8
and Figure 9
In Figure 8
, the trend of the end-to-end delay for a small transmission range (e.g.,
m) is increasing with respect to the network size in contrast to the end-to-end delay behavior resulted by large transmission ranges. The main arguments behind this behavior are: (a) in case of a small
, the packet fails to reach D
if it is relayed over multiple hops, due to channel impairments, path losses, high error rate of the acoustic channel, and so forth; (b) because of the network sparsity, end-to-end connectivity could not be established. Therefore, in a sparse network scenario, only the short hop-distant packets can reach D
and experience a small end-to-end delay. On the contrary, in a dense network environment, the packet has to traverse a large number of hops, which causes a very long end-to-end delay. The similar trend is also observed in Figure 8
. Furthermore, it is also noticed in Figure 8
increases, the end-to-end delay descends, which is due to the collision that leads to a large data packet loss. This collision happens when different potential forwarders relay the same data packet because of the small holding time difference and the large propagation delay between these forwarders, refer the theory related to Figure 2
. Data packets in the proposed ESEVBF scheme experience about
less end-to-end delay compared to AHH-VBF in any network size scenario with
m, respectively. As the
increases, the end-to-end delay of both the schemes becomes identical because the Data packet is forwarded through less number of hops and directly reaches the sink node. The overall performance gain (percent improvement in end-to-end delay) achieved by the proposed ESEVBF is shown in Figure 10
After the data packet broadcast and end-to-end delay analysis, we investigate the overall energy consumption in the network. Figure 11
and Figure 12
show the total energy consumption in the network during the whole simulation duration for different network size and transmission range, respectively. In underwater acoustic networks, transmission of a packet is the most energy consuming operation in the network compared to the packet reception, idle listening, sensing, and the processing operations. As data packets are large in size compared to the control packets, therefore, their contribution to the energy consumption dominates the energy consumed by the transmission of other packets or network operations. Therefore, the trends of the overall network energy consumption graphs are comparatively similar to the results that depict the total number of forwarded copies of the data packets in the network. Hence, the main reasons and arguments related to energy saving are similar to the one that avoids broadcast of more copies of the data packet. As a result, the overall energy consumption of the proposed scheme is less than the AHH-VBF.
The simulation results show that the energy consumption in the sparse network is very small because most of the data packets could not be forwarded further in the upstream direction towards the Sink. Similarly, the number of forwarders, as well as the data packet receiving nodes (receiving energy consumption), are very few, which is one of the reasons for this small network energy consumption. Conversely, the opposite is the case for a dense network scenario where the successful communication of data packets increase energy consumption in the network. As the number of data sources are fixed in all the simulation scenarios, hence, we also recorded the overall network energy consumption after the individual broadcast of each data packet by the source node, refer Figure 13
. The results show that maximum energy maximum energy saved by the proposed scheme is approximately
. The average percentage less energy consumed by the proposed scheme in comparison to AHH-VBF is summarized in Table 1
As we already discussed that more energy is consumed by the underwater acoustic networks, which can deplete the battery power of some nodes (dead nodes) during the simulation duration. Therefore, we also studied the number of dead nodes during the simulation period, as shown in Figure 14
. It is evident from the figure that the battery power of few nodes is completely consumed in the network scenario of 200 nodes and
m, Figure 14
a, because a small number of data traffic is handled by the network. In the similar network scenario but for the large
, more network nodes die out, because more data packets are communicated in the network. The battery power of more number of nodes deplete when the network becomes denser. However, it can easily be seen from the results that for
m, the number of dead nodes is larger than the
and 500 m, refer Figure 14
a. The obvious phenomenon behind this is that in an extremely dense network scenario, a large number of potential forwarders get the chance to forward data messages because of a very small difference between their holding time. Therefore, the collision probability at the next hop increases and the next hop nodes fail to further communicate the data packet. Hence, the data packet communication is restricted only to a few hops in the network. On the other hand, in
m, more data packets are successfully propagated in the network, which consumes more network energy and results in a large number of dead nodes in the network, refer Figure 14
c,d for further details. Similarly, in the dense network scenario, e.g., Figure 14
d, the small transmission range (e.g.,
m) consumes more energy due to the fact that most of the nodes participate in the Data packet forwarding. Hence, large number of nodes die out for small
m compared to the large
m. From the results, it can easily be seen that the battery power of a very small number of nodes is depleted during the simulation of ESEVBF. Resultantly, it increases the underwater monitoring duration and the network with ESEVBF can survive for a longer duration that the one using AHH-VBF.
From the above analysis, it is clear that the proposed scheme saves energy, reduces broadcast storm, and communicates data packets with less delay. In addition to these performance gains, the average number of hops that each data packet traverse from S
is also analyzed for different network sizes and
, refer Figure 15
and Figure 16
, respectively. The hop count data coincides with our previous analysis that for a small
and network of 200 nodes, the average number of hops is smaller than the large
s. The rationale for this behavior is that the data messages from sources that are at distant location, fail to reach D
due to unavailability of the path(s). However, when the network becomes dense, the data packet has to traverse many hops to reach D
. On the other hand, it can easily be noticed from the results that the average number of hop counts of the proposed scheme is smaller than the AHH-VBF because the forwarder selection criteria mostly considers the node’s closeness towards the center of the virtual cylinder and the Sink node. Notwithstanding, the proposed scheme scales the holding time with residual energy factor plus the expanded ratio of the closeness towards the center of the virtual cylinder. This may increase the chances of other nodes to serve as potential forwarders that have large residual energy and slightly different virtual cylinder distance ratio. This is the reason that for large
, the average number of hops traversed by the Data packet in ESEVBF is slightly larger or almost identical to the AHH-VBF. The impact of this slightly larger hop count is reduced by taking the advantage of utilizing the neighborhood information in holding time calculation and reducing the holding time of the potential forwarder that has a smallest holding time among neighbors. Therefore, this marginally high hop count factor has not much impact in presence of the other significant performance gains achieved by the proposed scheme.
Along with those performance gains, finally, we simulated the packet delivery ratio or PDR
. PDR is one of the fundamental performance measures of every routing strategy is also analyzed. Figure 17
and Figure 18
show the PDR comparison of the proposed and the AHH-VBF scheme for different network size and transmission range, respectively. Figures show that the proposed scheme has almost similar PDR as the AHH-VBF as the network size or
is very large. On the other hand, the ESEVBF has slightly less PDR because ESEVBF selects the potential forwarders in the in PFZ that have closeness to the center of the cylinder and have larger residual energy. However, due to small number of nodes in the PFZ due to network sparseness, it is quite difficult to find suitable forwarders. Hence, the PDR of the ESEVBF is lower in those network scenarios. Conversely, in the dense network scenario, the chances of finding the suitable forwarder from the PFZ becomes very high that increases the PDR. The above results show that the proposed scheme achieves energy efficiency, broadcast less number of Data packets, and has lower end-to-end delay at the cost of slightly lower or almost identical PDR is different network scenarios.
In the next subsection, we briefly analyze and contrast the impact of Sink mobility on the performance gain of both ESEVBF and the AHH-VBF schemes.
4.3. Simulation Results in the Mobile Sink Scenario
In the previous section, we critically analyzed and reasoned about the performance gain of the ESEVBF in the network scenario with the static Sink. However, the literature suggests that introduction of the Sink mobility enhances the network performance in terms of small end-to-end delay, reduces the broadcast of the number of Data packets, minimizes energy consumption and increases the network lifetime, and so on. Although the AHH-VBF has not been tested in the mobile Sink network scenario, however, we extensively simulated AHH-VBF as well as our proposed ESEVBF in the mobile Sink paradigm and the results are discussed below.
shows the total number of copies processed within the network of varying size and transmission range. It is evident from the figure that the Sink mobility minimizes the broadcast of the number of data packet copies in the network. The reason behind this phenomenon is that Sink moves within the network and passes near the data generating and forwarding nodes. Hence, these nodes just have to forward data packet at fewer hops to reach the Sink that is within the close proximity of these nodes, refer Figure 20
that shows average number of hops the data packet traverses in the network to reach the Sink. Figure 19
a,b explicitly show that the ESEVBF and AHH-VBF forward far more less copies of data packet in the mobile Sink scenario as compared to the one with static Sink. The analysis shows that ESEVBF for any network size in a mobile Sink scenario processes
less number of data messages in the network compared to static Sink scenario for
of 500 m, 600 m, 700 m,⋯, and 900 m, respectively. Similar data packet gain is also achieved by the AHH-VBF in the network with Sink mobility. Additionally, it is also obvious from the figure that ESEVBF outperforms the AHH-VBF in any network scenario. The mobility enables the Sink to vary its proximity respective to the data packet source and the forwarder nodes. Hence, the number of hops traversed by the data packet is far more less in the mobile Sink network than the static Sink network scenario, refer Figure 20
Above analysis shows that the Sink mobility in the underwater network scenario significantly reduces the number of data message copies and the number hops the message traverses in the network. In result, it can easily be predicted that the end-to-end delay must also be alleviated in this network setting. Figure 21
shows the end-to-end delay experienced by the data message in a network with static and mobile Sink for varying (a) network size (b) transmission range. A notable difference in the end-to-end delay can be observed in the figure. On average, the ESEVBF lessens about
of end-to-end delay in the network with mobile sink than the network without sink mobility for the size of 200, 250, 300,⋯, and 450 nodes and all
. Notwithstanding, the AHH-VBF achieves
less end-to-end delay for the network of 200, 250, 300,⋯, and 450 nodes, which is closer to ESEVBF’s performance gain.
Next, the influence of Sink mobility on the overall network energy consumption is shown in Figure 22
. Results in that figure show the similar impact on the energy conservation as it has on the data copies processed in the network. It also verifies the claims in the literature that Sink mobility achieves energy efficiency in the network, which is about
more energy conserved than the static Sink scenario. Finally, the PDR improved by the proposed and conventional scheme is shown in Figure 23
. The overall PDR improved by the Sink mobility for ESEVBF and AHH-VBF Figure 24
a,b, respectively. From the above discussion, we can easily conclude that the Sink mobility improves the performance of the routing protocols.