1. Introduction
Underwater Wireless Sensor Networks (UWSNs) have attracted both academia and industry to explore the underwater resources by enabling a variety of aquatic applications. For instance, military defense, monitoring the aquatic environment, disaster prevention, pollution monitoring, underwater mineral extraction, etc. [
1]. The sensor nodes are randomly deployed over a specified geographic volume with the ability to sense, gather and transmit data towards the destined location (that may be a single sink or multiple sinks) [
2,
3].
Data communication in the acoustic medium faces several challenges due to the peculiar features of the aquatic environment like high propagation delays, high deployment cost, node movement due to water currents, energy constraints, limited bandwidth, etc. [
4,
5]. Various routing protocols are proposed to enhance the network lifetime with optimal energy consumption and minimize delay from the source to the destination using direct or multihop data transmission mechanisms [
6,
7,
8]. To deliver the data to the destination, geographic routing is widely used for both aforesaid data communication methods depending on the nature of the environment. Geographic routing uses the greedy forwarding strategy where each node finds the shortest path towards the destination to save its energy. However, in the greedy forwarding strategy, immutable forwarder node selection is inevitable, which leads to immature depletion of the node’s battery and creates a void hole [
9]. The void hole is avoided using the Adaptive Hop-by-Hop Vector-Based Forwarding (AHH-VBF) routing protocol [
1]. It uses a pipeline to restrict the transmission range, and it is adaptively adjusted to amend the forwarding area to reduce duplicate packets’ transmission.
However, the void occurrence is not avoided in sparse deployment using this algorithm [
9], because it reacts once the data packet is trapped and the data communication process is paused. This may occur because of the variation in the path quality, which is important for energy efficiency in IoT-enabled WSN. The major factors to be considered in the path to avoid void occurrence are: shortest distance and lesser number of links to enhance the network lifetime [
10,
11]. The IoT-enabled WSN has the ability to sense, gather and transmit a huge amount of data over a long distance. However, the limited batteries are the major hurdle in successful network operations. Thus, we need to schedule the data transmissions in energy constraint networks. Similarly, the network virtualization is another important aspect to find the path fault successfully in IoT-enabled WSN. The virtualization focuses on the optimal utilization of sensing resources. Moreover, it also supports diversity in the network and enables an efficient management of power resources. However, it has a reactive approach in handling the link failure [
12]. The IoT-enabled WSN has been helpful in connecting anything, anywhere. Anywhere means sensors, vehicles, cameras, watches, phones, etc. [
13]. The vehicles with customized sensors can allow communication with nearby IoT-enabled WSN. In fact, vehicles are feasible for communication because they do not have the energy limitation problem. However, our focus is on the energy constraint in IoT networks in which the device battery needs to be efficiently utilized. Thus, we need to have a routing algorithm that takes precautionary measures in advance to save data packet loss and handle the occurrence of a void node efficiently while saving the node’s energy [
14]. Therefore, to achieve energy efficiency along with minimum delay in IoT-enabled UWSNs, the need for robust routing algorithm emerges, which can be adapted according to the available resources.
In this paper, we propose four schemes: Adaptive transmission range-based WDFAD-Depth-Based Routing (DBR) (A-DBR) and Backward transmission-based WDFAD-DBR (B-DBR) are proposed to reduce the probability of void hole occurrence. While the A-DBR scheme adjusts its transmission range to overcome the void hole problem to continue the greedy forwarding of the data towards the sink, B-DBR exploits a fall back recovery mechanism to find out an alternative route for delivering data at the destination. Additionally, we propose Cluster-based WDFAD-DBR (C-DBR) to minimize end-to-end delay and reduce energy consumption. The Collision Avoidance-based WDFAD-DBR (CA-DBR) handles the collision problem by selecting a potential forwarder node with the minimum number of neighbors. The main scientific contributions of this paper are:
Two techniques, A-DBR and B-DBR, are proposed to avoid void holes.
Two techniques, CA-DBR and C-DBR, are proposed to avoid collision and minimize the packet drop ratio.
C-DBR selects the forwarder node with the maximum residual energy in order to enhance the lifetime of the network.
The proposed schemes are compared with WDFAD-DBR in terms of average packet delivery ratio, energy tax, end-to-end delay and accumulative propagation distance.
The rest of this paper is organized as follows: In
Section 2, related work on existing schemes in UWSNs and the problem statement are presented. In
Section 3, the background is discussed including the system model, energy consumption and propagation models.
Section 4 describes the proposed schemes in detail. Simulation results are presented in
Section 5. Performance trade-offs are given in
Section 6, followed by the conclusion in
Section 7.
2. Related Work and Problem Statement
The proliferation of sensing devices has enabled real-time monitoring through WSNs. These devices are low cost and can easily be deployed to gather data from the region of interest. In this perspective, UWSN has emerged to provide a feasible surveillance system for the rich resource of the acoustic environment. Therefore, the research community wants to explore underwater resources; however, an efficient routing algorithm that can provide reliable communication is desired, such as AHH-VBF, which is used to reduce energy consumption by adjusting the range of the forwarding vector [
1]. To reduce energy consumption and avoid void hole occurrence, it dynamically adjusts the transmission power at each hop along with the vector. However, with the adjustment in transmission power, the energy dissipation increases.
A GEographic and opportunistic routing with Depth Adjustment-based topology control for communication Recovery over void regions (GEDAR) is proposed in [
15]. It uses the greedy routing strategy to forward packets towards the sink node. Moreover, a priority is assigned to each neighbor node to avoid redundant transmissions by only allowing the highest priority node to transmit the data. In case the transmission fails, the node with low priority in the table resumes transmission from an alternate route. Additionally, this protocol uses a depth adjustment mechanism to provide continuous communication among the network nodes. However, moving nodes to a new depth causes excessive energy consumption and high end-to-end delay.
The Hydraulic-pressure-based anyCast (HydroCast) [
16] algorithm was designed to deliver data reliably to any sink positioned at the surface of the water. The forwarder node is chosen on the basis of the packet status and the cost of the link. Through a gauge, the depth information is obtained for successful data transmissions. This scheme has improved Packet Delivery Ratio (PDR) due to the low ratio of void node occurrence at the cost of high communication overhead, which causes more energy depletion of the network nodes.
In [
17], the authors proposed the Hop-by-Hop Dynamic Addressing-based Routing Protocol for Pipeline Monitoring (H2-DARP-PM). Dynamic hop addresses are assigned to each hop to enable efficient forwarder node selection. This scheme assigns dynamic a hop address to every node that contributes to data forwarding. This scheme improves the PDR; however, the energy consumption is considerably high.
Delay-sensitive schemes: Advancement of localization-free routing protocols of DBR, EEDBRand AMCTD [
18] are presented for time-critical applications. The authors have made these routing protocols scalable according to the application requirements to achieve minimum end-to-end delay along with the minimal energy dissipation. However, duplicate packets are forwarded very often because of the hidden terminal problem. In delay-sensitive EEDBR, the energy consumption is high, whereas the packet drop ratio is considerably improved in AMCTD.
Free Space Optical (FSO) and Electro Magnetic (EM) wave-based communication schemes [
19] have been used to examine an analytical framework to find an optimum range of clusters. Moreover, the logical results are computed to change the location of the sink to three different points: the center, corners and midpoint of the network field. This scheme results in less energy consumption at the cost of high end-to-end delay.
To save energy, sleep-awake scheduling is a widely-accepted mechanism. For instance, in [
20], the authors nominated an initiator node after the configuration of network nodes to gather data from the desired nodes. The communication phase begins with the initialization of the transmission phase. First of all, a head node is selected at each hop to lead the data packet towards the destination. Only the head node transmits the data, and nodes in the neighborhood are switched to sleep mode to avoid the unnecessary dissipation of the node’s battery. This scheme reduced the energy consumption significantly with enhanced lifetime and increased PDR. However, the immutable selection of the head node results in the sudden death of the node and degrades the network performance.
To collect data at distributed points, clustering is performed because it is scalable and flexible in nature. The same features motivate the research community to explore this area in more detail. In [
21], the network was divided into irregular clusters for making local routing decisions to avoid high data traffic at the sink node. Moreover, the algorithm forms irregular clusters based on a layered architecture for event coverage and obtains the expected value of the clusters through theoretical analysis. Additionally, this scheme uses a recovery strategy to balance the energy consumption among the clusters to enhance the performance of the network.
A Particle Swarm Optimization-based Energy-efficient Cluster Head Selection (PSO-ECHS) was proposed in [
22]. To balance the energy consumption, various control parameters were taken into the consideration like distance within the clusters and from the sink along with the residual energy of each node in the cluster. With the help of the aforesaid metrics, a weighting function was formulated, and probabilistic value was computed to nominate and rotate the head node for efficient energy consumption. This scheme achieves high PDR at the cost of delay.
In [
23], three schemes were proposed: Sparsity-Aware Energy-Efficient Clustering (SEEC), Circular SEEC (CSEEC) and Circular Depth-based SEEC (CDSEEC). In SEEC, two mobile sinks are deployed in sparse and dense regions to collect information and to reduce the probability of energy hole occurrence; while a different geometry is considered for CSEEC to analyze the mobility of sinks, which improves the PDR and maximizes the network lifetime. The same topology is considered for CDSEEC with a different mobility pattern. The trade-off occurring against energy efficiency and PDR is the highest end-to-end delay.
Depth-Based Routing (DBR) [
24] uses a greedy approach to deliver packets towards the sink based on the depth of a forwarder node. Each eligible source node transmits a packet based on depth and also calculates the holding time to avoid duplicate packets’ transmission among the network nodes. However, the consideration of only distance in the selection of the next hop node forces the immutable nomination of the forwarder node. This leads to sudden death of the intermediate nodes. Moreover, the holding time is not synchronized, resulting in transmission from the neighbor nodes before even the acknowledgment arrives. However, DBR benefits from high network lifetime and PDR at the cost of only end-to-end delay.
An improved Adaptive Mobility of Courier nodes in Threshold-optimized DBR (iAMCTD) [
25] is presented to handle flooding, latency and path loss. The routing is performed on demand to maximize the network lifetime through an optimized mobility pattern of courier nodes, whereas, the Energy-efficient Channel-Aware Routing Protocol (E-CARP) [
26] provides improved network lifetime and reduced energy consumption by the reactive routing approach.
In Adaptive Relay Chain Routing (ARCR) [
27], the authors introduced mobile sensor nodes to overcome the energy hole problem. Additionally, clusters were formed for collecting data via mobile nodes to improve the network performance. This routing mechanism achieved energy efficiency and maximum lifetime at the cost of low PDR.
The proposed work is different from the discussed related work based on the following distinguishing features. In order to reduce the probability of void hole occurrence, A-DBR adjusts its transmission range adaptively. However, the adjustment of transmission power causes extra energy consumption when the distance increases between the source and destination. Thus, B-DBR looks for a forwarder in all possible directions within its transmission range to find an alternate path to deliver data at the destination; while C-DBR and CA-DBR minimize end-to-end delay along with collision by making clusters in the network. The related work is summarized in
Table 1.
2.1. Problem Statement
To efficiently utilize the node battery and to reduce the end-to-end delay, the research community has been devoted to bringing improvement in routing algorithms designed for UWSNs. For instance, WDFAD-DBR [
4] considers only two metrics: the depth of current and next expected forwarder node. Although, the probability of void hole occurrence is reduced and inefficient energy consumption during nodes communication is minimized, the probability of void hole occurrence still exists, as illustrated in
Figure 1.
When the source node S initiates communication and finds in its communication range, before transmitting the data packet to , it acquires information about its neighbor node. It locates in its transmission range and delivers the data to . Thus, it acknowledges the S with non-void node status and receives the data packet. However, when looks for its neighbors, it finds , which has no further nodes in its transmission range, resulting in loss of the data packet. Thus, this process will continue until the death of . Additionally, this scheme is receiver based, where avoidance of duplicate packets is very difficult. The reason is that neighbors in the hidden terminal region are unable to receive the acknowledgment, leading to redundant transmissions at the destination. Moreover, it also leads to channel interference in the case of simultaneous transmissions over the acoustic wireless channel. Furthermore, it causes collision, leading to a high packet drop ratio and more end-to-end delay. To overcome the aforementioned problems, we propose four schemes: A-DBR, C-DBR, B-DBR and CA-DBR, to improve the network performance. The details are given in the following sections.