Diverse Congestion Control Schemes for Wireless Sensor Networks

Wireless Sensor Networks (WSNs) comprised of battery operated sensor nodes that collect data from their neighbor nodes and transmit the aggregated information to the sink node or the Base Station (BS). This may result in congestion near the BS and leads to a bottleneck situation in the network. In this paper, an extensive study of earlier reported diverse congestion techniques explicitly diverse Algorithm based - and Layer based-congestion techniques is carried out. Accordingly, a recommendation is drawn based upon their performance comparison. Furthermore, a demonstration is carried out for contemporary earlier reported strategies such as Pro-AODV, CC-AODV, EDAPR, ED-AODV and PCC-AODV by evaluating delay, packet delivery ratio (PDR) and packet loss ratio (PLR). Accordingly, a recommended congestion strategy is suggested depending upon the comparison of the demonstrated schemes.

due to the simultaneous transmission attempts executed by different sensor nodes. Congestion results in data loss and degrades the performance of a network by decreasing the throughput. Depending upon the impact of congestion, it is classified into two categories namely: local congestion and global congestion. Congestion at the local level mainly occurs either due to the node or link failure. However, global congestion occurs in the whole network and hence, minimizes the data transfer rate. In WSNs global congestion rarely occurs due to availability of the sufficient resources. However, the local congestion occurs more often due to the limited battery life and memory availability of deployed sensor nodes. Local congestion can occur either due to node failure or buffer overflow. Buffer overflow occurs if several nodes try to transmit data packets simultaneously to the single destination sensor node. This work investigates the earlier reported congestion techniques and has been classified into two categories named Algorithm based Congestion Techniques and Layer based Congestion Techniques. The remaining of the paper is organized as follows: Section 2 highlights the diverse congestion control techniques and Section 3 presents the simulation results of presently available congestion control techniques and conclusion is drawn based upon the comparative analysis of diverse congestion control techniques and mentioned in section 4.

Diverse Congestion Control Strategies
In the layer based congestion technique, each sensor node senses the channel for a fixed amount of time before transmitting its own data. This waiting time leads to significant performance degradation and can be reduced by introducing random sensing times instead of a fixed waiting time with usage of sequential detectors. This will also minimize the average number of samples required to make a decision for transmission of data onto the channel and hence leading to reduction in the communication delay, power consumption and the risk of collisions [12]. Neural Network based congestion detection (NNBCD) a layer based congestion protocol has been proposed to detect the level of congestion during the significant packet drop in the network by considering the buffer occupancy and traffic rate [13]. Further, a Delay-aware congestion control (DACC) protocol estimates channel occupancy based on buffer occupancy and transmission time of packets. REDD (Real-time and Energy-aware Directed Diffusion) routing protocol improves the performance of DACC protocol by reducing the estimation time and improves the network efficiency [14]. Further, an intelligent collision avoidance mechanism is proposed for the vehicular ad hoc network that uses ultrasonic sensors and IR sensors for front barrier detection and controls the flooding effect occurring due to duplicate message transmission [15].Energy based collision avoidance (ECA-MAC) technique handles WSN energy constraints and the QoS designed for versatile applications [16]. Based on different energy levels of nodes, ECA-MAC protocol controls the access to the channel and reduces the collision using different contention windows and hence improves the lifetime of the network. Modified Opportunistic Routing Protocol (MORR) improves the data routing in WSNs by allowing only the sensor nodes that have higher residual energy and minimum neighbor nodes to transmit the data packets to reduce the congestion in the network [17]. Table 1 shows the comparison of the layer based congestion control mechanisms for collision detection. Congestion control for MUlti-class Traffic (COMUT), a cluster-based mechanism supports multiple classes of traffic for sensor networks [18]. COMUT organizes the whole network into small, independent parts, and each part looks for congestion within its localized scope. This result in abating congestion and reducing the wasteful packet drops to save energy; however, COMUT fails to estimate the network traffic intensity. Congestion control for Sink to Sensors (CONSISE) is based upon co-activities like receiving, transmitting, transmit-decision and receive-decision [19]. This slows down the communication as the quantity of sensor nodes and the generated data traffic increases. To overcome this problem and to achieve high throughput, an adaptive duty-cycle based congestion control (ADCC) scheme had been proposed [20] that control both resource and traffic inside the network. The ADCC scheme reduces the control packet overhead by increasing and reducing the packet reception-/transmission-rate of the receiving/sending node respectively. Further, the ADCC scheme, a light weight congestion control the energy efficient scheme uses the duty-cycle adjustment but is not appropriate for Many-to-one sensor networks. Thus, two distributed algorithms had been proposed to efficiently find an alternative path using the neighboring table and mitigating the congestion by using an idle path instead of restricting transmit rate [21].
To improve the network lifetime, an On-demand node placement strategy reacts to congestion regions and transmits the traffic to non-congestion regions [22]. However, buffer overflow results in high packet loss and a large queuing delay that further augments the packet service-time and energy wastage. An early congestion detection and an adaptive routing (EDAPR) overcomes this drawback [23] by constructing non-congestedneighbors (NHN) list and discovers the path from the source node to the destination node using NHN node. Each sensor node detects the congestion occurrence or its probability and sends a warning message to NHN sensor nodes to discover an alternative path to reach to the destination node. This improves network lifetime but at the cost of high packet loss and prolonged delay [24]. Further, an optimized fuzzy logic based congestion control mechanism improves the performance of the network that considers buffer occupancy and congestion index (CI) and estimates the congestion occurrence but at the cost of increased network load [25]. A cross-layer congestion control (CLCC) strategy overcomes this limitation by considering rate adjustment for controlling the congestion and application-oriented design for the different situations but at the cost of long delay [26][27][28][29][30][31][32][33]. Further, an adaptive congestion control protocol (ACCP) improves the network performance by using a sink switching algorithm to switch between traffic congestion control logic (DelStatic) and resource congestion control logic (DSR) [34]. The packet loss or drop rate can be minimized by adopting a priority based congestion detection and the avoidance scheme. This strategy differentiated data delivery during congestion which includes the packet priority assignment, priority based queue scheduler and dynamic dual-path congestion algorithm [35]. Apart from this, still there are chances of the packet dropping that can be overcome by cross-layer congestion Control algorithm applicable at both node-level and link-level. This improves the network lifetime by compressing the transmission signal and allocated the channel that makes high levels of data flow but at the cost of seizing the same channel while maintaining a high level of data flow [36]. This can be overcome by using feedback congestion control which is based on linear discrete time control concept that improves the system performance in terms of energy consumption and high throughput, but is affected by buffer overflow issues that causes significant drop in packets [37]. A pro-active protocol uses information from the AODV routing table to minimize this packet drop rate and congestion. The drawback of this technique is whenever a node decides to broadcast a route request (RREQ); it only chooses a subset of nodes listed in the table [38]. To overcome this, a cross-layer admission control (CLAC) mechanism is reported which is based on technique that preview packet end-to-end delay and forward the same if expected to meet end-to-end delay, otherwise dropped the packets. This enhances the performance in terms energy efficiency but at the cost of high packet loss [39]. This further is a combat with a congestion control AODV algorithm that is used to avoid link break and in which the received signal strength was being used as cross layer design parameter. The only drawback of this algorithm is drop of packets whenever the required transmission power and the distance between the nodes increase [40]. Traffic redirection based congestion control transport protocol (TRCCTP) is reported which is based on finding the optimal path that redirects the traffic from congested area to uncongested area and minimizes the packet drop ratio [41][42][43]. Another technique named congestion detection and alleviation (CDA) mechanism is reported which is based on carrier sense multiple access with collision avoidance (CSMA/CA) in which node level congestion and link level congestion is detected by buffer utilization and congestion is alleviated by rerouting the traffic to a less congested route, which minimizes the packet drop ratio [44]. Another approach known as congestion avoidance and congestion control mechanism was proposed to overcome the problem of packet dropping. In congestion avoidance, the node uses the lower hierarchy nodes to forward the data and congestion control mechanism will detect the congestion using queue length. The only drawback of this approach is buffer overflow [45][46]. Congestion control [47][48] adaptive routing overcomes the problem of buffer overflow. In this congestion detection, alternate path computation and rerouting of packets are being carried out based on free space available in the buffer, available bandwidth, hop distance and residual energy on the path [49]. Another proposed algorithm is initial constant congestion window that calculates the values of threshold on the basis of acknowledgements, that improves the lifetime of sensor nodes by introducing a novel mobile node as a charger and handles the congestion of bottleneck links by using initial constant congestion window (ICCW) [50]. Table 2 shows the comparison of various Algorithm based congestion control mechanisms.

Simulation Results of Congestion Control Strategies
Furthermore, the performance of different algorithm based congestion techniques via different metrics; for instance, packet delivery ratio (PDR), packet loss ratio (PLR), Throughput and delay at different node density are computed. Table 3 shows the comparison of existing congestion control strategies like PRO-AODV (Proactive), CC-AODV (Congestion control), EDAPR (Early congestion detection and adaptive routing), PCC-AODV (Predictive congestion control routing) and ED-AODV (Early congestion detection control routing) in terms of number of flows.  Figure 1-3 shows the comparison of End to end delay, packet delivery ratio (PDR) and Throughput for Pro-AODV [38], CC-AODV [40], EDAPR (Early congestion detection and adaptive routing) [23], ED-AODV (Early congestion detection control routing) [23] and PCC-AODV (Predictive congestion control routing) [24] protocols respectively.     Table 4 shows the comparison of different congestion control strategies with varying number of nodes for PRO-AODV (Proactive), CC-AODV (Congestion control), EDAPR (Early congestion detection and adaptive

Research Article
Vol. 12 No.6 (2021), 2380-2389 routing), PCC-AODV (Predictive congestion control routing) and ED-AODV (Early congestion detection control routing) in terms of Packet delivery ratio (PDR), Packet loss ratio (PLR) and Throughput.   [38], CC-AODV [40], EDAPR (Early congestion detection and adaptive routing) [23], ED-AODV (Early congestion detection control routing) [23] and PCC-AODV (Predictive congestion control routing) [24] protocols respectively. From Fig 4 and Fig 5, it has been demonstrated that PCC-AODV outperforms in terms of packet delivery ratio, as with increase in number of nodes, routes are efficiently managed in this mechanism and so does the packet loss ratio which depends upon PDR because routes are being managed using optimum route maintenance policy in this mechanism. Fig 6 describes the CC-AODV outperforms than other mechanisms in terms of throughput (233 at 50 number of nodes) because it aims to reduce performance degradation that occurred due to congestion.   The outcomes of the demonstrated diverse congestion techniques show that PCC-AODV outperforms the other algorithms in terms of PDR and PLR. Further, it has been observed that ED-AODV claims a decreased approach in dense traffic rate.

Conclusions
Controlling congestion is a demanding and complex issue in the WSN networks having limited resources. The numerous strategies have been discussed previously to manage congestion to expand the network lifetime by using restricted available resources adequately. Researchers have categorized these schemes mainly into algorithm-and layer-based congestion control algorithms. This article examines earlier reported Algorithm-and Layer-based strategies, present a comparative and discussed their pro and cons accordingly. This may help the researchers to opt for an appropriate congestion strategy as per network design and required QoS services. Furthermore, the authors assessed current existing congestion control techniques in terms of delay, throughput, PDR and PLR and claimed PCC-AODV as the recommended congestion strategy among the other demonstrated strategies in this work.