Cognitive radio sensor networks (CRSNs) are the next generation wireless sensor networks (WSNs) that mitigate overcrowded unlicensed spectrum bands by opportunistically using temporally unoccupied unlicensed and licensed spectrum bands. In this paper, we propose a new energy- and cognitive-radio-aware routing (ECR) protocol that addresses the unique challenges in CRSNs, including dynamic spectrum access, single transceiver, and energy constraint. In particular, our proposed routing protocol performs joint node-channel assignment by taking energy into consideration, is aware of cognitive radio at the network layer, and can seize spectrum opportunity in other spectrum bands. We present a simple analytical model of the proposed ECR in the viewpoint of network-wide energy and compare it with that of the ad hoc on-demand distance vector (AODV) routing protocol. Furthermore, our simulation results show that, in relatively heavy traffic environment, ECR outperforms AODV in terms of network lifetime and packet delivery ratio. Nevertheless, scalability and communication complexity become the major issues of this protocol.
Wireless sensor networks (WSNs) are a special case of ad hoc networks and have been widely used for monitoring physical phenomena, such as human activities and environment monitoring. Since it is intended to be easily embedded in the physical environment, WSNs are designed with minimum computational facilities and limited power resources.
In the future, however, ISM (Industrial, Scientific, and Medical) bands are projected to be congested and overloaded due to numerous wireless networks utilizing the same bands [
Meanwhile, cognitive radio ad hoc network (CRAHN) technology offers a good solution to increase spectrum utilization by making use of temporally unused spectrums in an opportunistic manner. By combining cognitive radio (CR) capability to WSNs, the spatially overlapping wireless networks may coexist in ISM bands with minimum interference, and spectrum utilization can be increased. Hence, cognitive radio sensor networks (CRSNs) [
Opportunistic usage of the lower frequency of the licensed bands offers a number of benefits to CRSNs. The main advantage is that lower frequency has better propagation characteristics than the unlicensed 2.4 GHz frequency, such as longer transmission range and better penetration to obstacles. In [
From the view point of network layer, CRSNs have two fundamental issues: joint node-channel assignment for enabling dynamic spectrum access (DSA) and energy consumption in hardware constrained networks. Nevertheless, the existing spectrum-aware techniques of CRAHNs and energy efficient techniques of WSNs cannot be directly applied to CRSNs.
There are studies aimed at addressing routing issues in CRAHNs as surveyed in [
For those reasons, there is a need for research on achieving an applicable routing protocols for CRSNs that combines the joint node-channel assignment of CRAHNs and the energy efficient techniques of WSNs. In this paper, we propose a novel routing algorithm called energy- and cognitive radio-aware routing (ECR), which takes advantage of both CRAHNs and WSNs to satisfy the routing requirements of CRSNs. The contributions of this paper are twofold. First, we introduce a novel routing protocol that meets the unique requirements of CRSNs. Second, we consider the energy and cognitive radio awareness at the network layer in CRSNs. Our extensive simulation results show that, in a cognitive radio-enabled networks with relatively heavy traffic, energy and cognitive radio awareness in our proposed ECR prolongs the network lifetime and improves packet delivery ratio.
The rest of this paper is organized as follows. In Section
We consider a typical sensor network, which comprises reduced function devices (RFDs) and full-function devices (FFDs), as per device classification in the IEEE 802.15.4 standard. RFDs are sensor nodes that are intended to perform extremely simple tasks and do not have routing capability. On the other hand, FFDs are high capability nodes which can operate as either personal area network (PAN) coordinator nodes, coordinator nodes, or sensor nodes. They can perform complex tasks such as routing and data aggregation. When there is data to be sent over network, the coordinator nodes collect the data from their sensor nodes and send the aggregated data to the PAN coordinator in multi-hop manner according to the routing protocol. From hardware’s point of view, each nodes is equipped with single half-duplex transceiver. Also, each node is battery-powered, except the PAN coordinator node which is the only main-powered device in the network.
Figure
Schematic of network model.
Since conventional WSN has no capability to utilize unused spectrum in other channels, it operates statically on a particular channel of unlicensed spectrum band. On the other hand, the CRSN can operate dynamically on multiple channels of 2.4 GHz unlicensed ISM bands and 680 MHz UHF licensed TV bands. As CRSN is designed to coexist with the legacy networks, the user priority concept is applied, where primary users (PUs) has higher priority than the secondary users (SUs) to utilize the channel. In this paper, we define PUs as the devices of both conventional WSN in unlicensed spectrum band and any networks in licensed band, and SUs as the CR-enabled devices of CRSN. That being said, CRSN is expected to improve the system throughput while maintaining coexistence with the conventional networks.
In this paper, we take channel heterogeneity issue into account by considering 16 channels in unlicensed spectrum band and 1 channel in licensed spectrum band, each of which has 2 MHz bandwidth and supports data rate up to 250 kbps. In addition, we dedicate 1 channel out of 16 channels in unlicensed spectrum band as the common control channel of CRSN.
In the following section, we explain the proposed energy- and cognitive-radio-aware routing (ECR) protocol. In principle, it enhances the ad hoc on-demand distance vector (AODV) protocol [
In general, the route discovery process of the proposed routing protocol is divided into four steps: route request, route selection, route reply, and route maintenance. The following subsections explain each of these steps.
As in the standard AODV protocol, route request (RREQ) packet is used to find any possible routes from source to destination node. Namely, the source node appends node information, and broadcasts RREQ packets to the neighboring nodes.
Before rebroadcasting a RREQ packet to its neighboring nodes, the intermediate node should examine two requirements. Firstly, it checks its own residual energy status. If the remaining energy is below a certain threshold level, then the intermediate node should drop the RREQ packet. Secondly, the intermediate node checks whether or not it has any common channels with the previous node. If there is no common channel between them, then the intermediate node should also drop the RREQ packet. Note that dropping RREQ packet is the same as withdrawing from the route discovery process. The former is intended to save excessive energy consumption of the node and to ensure that the load of network will be distributed more evenly over the network, while the latter is to ensure that at least one feasible route can be established between source node and destination node. Intermediate nodes which is able to participate in the route discovery process piggyback their node information to the RREQ packet. Afterwards, they rebroadcast the RREQ packet to its neighboring nodes and setup reverse path in their routing table as in the AODV protocol.
This procedure is done in every intermediate node until the RREQ packet reaches the destination node. In other words, the length of RREQ packet would increase as the number of hops increases. The node information appended to a RREQ packet on each hop is explained as follows. Node ID, which is used as a node identifier in the network. Channel availability information, which consists of licensed channel (L) and unlicensed channels (UL). Each channel is represented by one bit information, whose availability and unavailability are encoded by bit “1” and “0”, respectively. Node energy, which represents the ratio of residual energy and full energy of the node.
The format of new RREQ packet is shown in Figure
RREQ message format in ECR.
Upon receiving the first RREQ packet, the destination node (i.e., the sink node) starts the timer and waits for another RREQ packets in order to collect more route candidates. After the timer has timed out or the minimum number of alternative routes has been collected, the destination node will populate as many joint node-channel combinations as possible from the information in each RREQ packet. Then, the destination node selects the best route by considering the following factors.
The above criteria are formulized in the following route cost function, which is mathematically stated as:
After selecting the best route based on the cost function value, the destination node will assign the operating channel to each node involved in the corresponding route. Namely, the destination node unicasts the route reply (RREP) packet toward the source node, which contains joint node-channel assignment of every nodes involved in the route, including the source and destination node. The format of new RREP packet is shown in Figure
RREP message format in ECR.
Note that the joint node-channel assignment information is appended to the AODV’s RREP packet, where the channel number between two node IDs indicates the assigned communicating channel used by the node pair to transfer the MAC frame. Upon receiving an RREP message, the intermediate node will read the corresponding channel assignment and store it in its routing table. Afterwards, the RREP packet is forwarded to the next intermediate node until it reaches the source node.
The local repair mechanism of AODV protocol is used as the route maintenance strategy. Namely, whenever a link breakage occurs due to intermediate node’s failure or PU arrival, the intermediate node will first determine its position with respect to the source node and destination node. If it is closer to the destination node, then the intermediate node will broadcast local RREQ packets to search for local alternative path in order to bypass the failure node within a small number of hops. In the meantime, incoming data packets are buffered in the intermediate node during the local repair process. If the alternative path is found, then the original route can be used again. Otherwise, the intermediate node will broadcast the regular route error (RERR) packet to inform the source and destination node that the current route is broken and the source node should start a new route discovery process all over again.
In this section, we present a simple analytical model of AODV protocol and our proposed ECR protocol to calculate network-wide energy consumption metric. In general, our analysis is divided into two parts: route discovery and data transmission. Namely, we calculate the total energy consumed at the network for exchanging control packets to find the route, and for transmitting data packets once the route has been established, respectively. Since the names of the variables in the formulas may not be unique, we are going to explicitly mention which protocol they refer to while explaining the formulas.
Suppose that there are
Let
We also consider the energy consumption to retransmit the control packets during the route discovery process
Thus, the energy consumption in route discovery process
Let one session of data transmission be defined as the flow of data from the source node to the destination node starting from the beginning of data transmission until it is over. Then, the energy consumption to send data in a session
Since data packets are transmitted in the same channel as the route discovery process, there is a chance that collision may happen in the data packets. Thus, using the same value of
Thus, the energy consumption to send data in a session
Taking the worst case scenario when there is
While broadcasting RREQ packet, source node appends node information to the original RREQ packet of the AODV protocol. Hence the RREQ packet size increases constantly as it propagates through the network. The total number of bits of the growing RREQ packet
Thus, for the same square grid topology, the energy consumption to send the RREQ packet from source node to destination node
Assuming that each node which is involved in route discovery process has to sense the spectrum band once to get the current status of all channels before appending node information, the energy consumption for spectrum sensing
Let
Therefore, the total energy consumption in route discovery process in ECR,
Since control packets and data packets in the ECR protocol may be sent in different channels, we define the energy consumption to send the data packets over the network in terms of probabilities.
Let
Suppose that the destination node decides to use other channels than the common control channel for data transmission, the ECR protocol prefers using the licensed channel to the unlicensed channels if possible. Then, this preference is mathematically expressed as follows:
When data transmission is not using common control channel, there is energy consumption for switching from common control channel to the destined channel, and vise versa. Let
Thus, the energy consumption to send data packets on other channels than common control channel
For simplicity, we assume that data transmission in common control channel is subject to collision. On the other hand, there is no collision happened if data transmission is done on other channels than the common control channel, because the channels have been sensed by the nodes and we do not consider the channel obsolesence issue in this paper. Thus, the energy consumption to retransmit the data packet
Thus, the energy consumption to send data in a session
Taking the worst case scenario when there is
In this section, we compare the performance of the proposed ECR protocol with that of AODV protocol in terms of energy consumption metric. The simulation was done on both Matlab and ns-2 network simulator version 2.31 with cognitive radio cognitive network (CRCN) patch [
Default parameters for Matlab simulation.
Parameter | Value |
---|---|
|
31.32 mW |
|
35.28 mW |
|
103 mW |
|
35.28 mW |
Number of nodes | 25 |
Data packet size | 512 Bytes |
Packet rate | 60 packets/session |
Number of sessions | 24 sessions |
|
0.25 |
|
0.3 |
|
0.7 |
Default parameters for ns-2 simulation.
Parameter | Value |
---|---|
|
31.32 mW |
|
35.28 mW |
|
103 mW |
|
35.28 mW |
Number of nodes | 25 |
Initial energy | 5,000 Joules |
Network area | 1,000 × 1,000 m2 |
Packet rate | 16 packets/second |
Number of sessions | 20 sessions |
PU channel occupancy | 25% |
Simulation time | 1,000 seconds |
We consider a CRSN network which consists of 25 nodes which are arranged in a
We are interested in observing the effect of data size and the number of nodes to the network-wide energy consumption by means of the analytical model of energy consumption.
Figure
Network-wide energy consumption in various data size.
Figure
Network-wide energy consumption in various numbers of nodes.
Network lifetime is an important performance metric in sensor networks. We are interested in observing the network lifetime as we vary the PU channel occupancy level in the default operating channel by means of ns-2 simulation. We assume that around 70% of other unlicensed and licensed channels are statically available.
First we set the energy of each node in CRSN (except the sink node) to 5 Joule, and we measure the lifetime of the network by counting the number of alive nodes remaining over the simulation time. We observe the network lifetime under various numbers of data transmission sessions (10, 20, and 30 sessions) on the network while maintaining the packet rate and PU channel occupancy level at the default values. From all three cases shown by Figure
The number of alive nodes for the different number of sessions.
10 sessions
20 sessions
30 sessions
We plotted the number of cumulative packets received by the sink node and the packet delivery ratio (PDR) in Figures
The total number of packets received by the sink node.
Packet delivery ratio.
We try to figure out the system throughput of ECR protocol. We vary the PU channel occupancy level from 0 to 90%. As shown in Figure
System throughput.
PU channel occupancy of 0–90%
PU channel occupancy of 70–90%
As the PU channel occupancy level increases, both protocols suffer from performance degradation in terms of system throughput. Nevertheless, from 80% traffic load level and above, in our simulation AODV was not able to deliver any packets across the channel due to the collision of both control and data packets in the default operating channel. On the other hand, ECR was still able to deliver data packets by opportunistically exchanging them in the other channels. Therefore, the system throughput of the ECR protocol outperforms that of the standard AODV protocol by at least 25% in every level of PU channel occupancy.
Complexity of an algorithm is informally defined as how much resource it needs to accomplish its goals. In this subsection, we compare the time and communication complexity of AODV and ECR to perform route discovery and postfailure procedure.
Let
Communication complexity for route discovery is the communication overhead needed to establish a route from source to destination, and that for postfailure is the communication overhead to reestablish the route if the route fails. In the AODV protocol, the RREQ packet is straightforward and its packet size is fixed. Thus, the communication complexity of AODV for route discovery is
In this paper, we have proposed an energy- and cognitive-radio-aware routing protocol called ECR to address the unique challenges of CRSN, namely dynamic spectrum access, single transceiver, and energy constraint. To the best of our knowledge, there is currently no specific DSA-enabled routing protocol designed for low-rate wireless sensor networks (WSNs). In addition, in designing the proposed routing protocol for CRSNs, we take the energy and cognitive radio awareness into consideration at the network layer.
We compare the performance of our proposed ECR protocol with the industry standard AODV protocol. Our analytical model of energy consumption shows that the ECR protocol has better energy efficiency while transmitting larger data. Also, the ns-2 simulation results show that the ECR protocol achieves better performance in terms of network lifetime, packet delivery ratio, and system throughput in relatively high traffic network environment. Nevertheless, the ECR protocol has drawbacks in that it is not designed for scalability and it has high communication complexity.
The ECR features can also be implemented as an extension to the AODV protocol, which can be enabled or disabled network-wide according to the network traffic level. In this case, the design of cognitive engine would be of great importance to sense the network traffic changes and decide when to enable the ECR features. Hence, the focus of our future works will be on the integration of ECR features to the AODV protocol and the development of cognitive engine to sense the network traffic changes. Another future work is that the route cost can be developed more to adjust to the network condition. As for example, the weight of the metrics can be adjusted for the purpose of traffic engineering.
This work was supported in part by Research Funds from Chosun University, and it was also supported in part by the (The Ministry of Knowledge Economy) MKE, Korea, under the (Information Technology Research Center) ITRC support program supervised by the (National IT Industry Promotion Agency) NIPA (NIPA-2012-H0301-12-2008). The authors would like to thank the editor and anonymous referees for their helpful comments on this paper.