Energy,Traffic Load,and Link Quality Aware Ad Hoc Routing Protocol for Wireless Sensor Network Based Smart Metering Infrastructure

Abstract

Electricity industry is in the midst of revolutionary transition from outdated ageing power infrastructure to an intelligent sophisticated smart grid network utilizing modern communication technologies to enhance power generation, transmission, distribution, and consumption. Smart metering infrastructure is an integral part of the smart power grid revolution. Smart meters, in addition to their primary billing functions, serve as distributed Wireless Sensor Network (WSN) nodes for enhancing grid reliability. Existing ad hoc routing protocols are based on single routing criterion such as hop count. This single routing metric approach can overload and deplete resource constrained smart meters along preferred paths. A protocol is needed which is aware of energy level and traffic congestion of smart metering nodes. In addition, protocol should select route based on link quality for optimal routing. In this paper, a novel WSN ad hoc routing protocol ETL-AODV is proposed for reliable and energy efficient communication of smart metering nodes. Three simulation based case studies are conducted to analyze the performance of the proposed protocol, and relative comparison is provided based on four metrics: (i) packet delivery ratio (PDR), (ii) normalized routing load (NRL) (iii) average energy consumption and (iv) average end-to-end delay.

1. Introduction

This century has witnessed exponential increase in electricity consumption, and its unchecked use has brought us to the verge of a power crisis. Developed nations of the world are working on solutions for energy conservation. Smart grid is a revolutionary concept aimed at reducing energy wastages and making ageing electric infrastructure more efficient, reliable, and intelligent. The core component of smart grid is smart metering infrastructure which upgrades electric meters into smart communicating nodes relaying power consumption statistics and event reports to the power grid control centre in real time manner [1]. This has opened new horizons of research areas emphasizing on communication and networking of smart meters.

Smart metering infrastructure, also known as “neighbourhood area network,” enables two-way communications between consumers and electricity supplying companies. Smart meters communicate with home appliances and transmit their power usage statistics to grid control centre in regular intervals. This information is very critical for electricity suppliers as it is used for efficient power generation and distribution. Electricity supplying companies can have real time view of load/demand and can preemptively bolster the grid against interruptions leading to improved reliability in all stages of its operation. Both wired and wireless communication mediums are being utilized for smart metering deployment. Wired medium uses existing power lines as means of communication. However, it suffers from low data rate issues, frequent harmonics disturbances, and unavailability during powering failures. Wireless communication is being considered a key enabler technology for smart metering communications. One of the most attractive characteristics of wireless medium is absence of physical connection between nodes thus requiring minimal deployment cost and management. This ensures continued connectivity even in case of power failures.

Recently, wireless sensor network (WSN) is being rapidly utilized for interconnectivity of smart meters [2]. WSN is a class of ad hoc network which organizes the smart meters in an infrastructure less distributed reconfigurable topology. Smart meters with their wireless capability transmit their own and relay other meter readings to the power grid control centre. Many WSN ad hoc routing protocols have been proposed which consider the limited battery life and computational capability of sensor nodes. Data centric protocols reduce redundancy in data transmissions for prolonging nodes lifetime. They utilize Query method which is not suitable for time driven monitoring applications. Hierarchical routing protocols improve energy consumption by locally grouping the nodes into clusters. However, they require frequent exchange of control messages for cluster formation. Location aided protocols, although energy efficient, require the use of expensive localization hardware for their working. Most of the WSN routing protocols work on flooding mechanism for event reporting to the sink. However, smart metering infrastructure represents a specific kind of network deployment due to large number of static metering nodes communicating in outdoor environment in time driven periodic fashion. Most of the time smart meter reading collection exhibits a many-to-one scenario with all metering nodes communicating with the sink and overloading the nodes close to it. Furthermore, smart meters running on power lines use rechargeable batteries for wireless communication module and back up requirements in case of power failure scenarios so energy consumption of protocols should also be considered.

Ad hoc on demand distance vector (AODV) [3] routing protocol is utilized by the Zigbee standard for deployment of WSN of home sensors and appliances. In AODV, nodes broadcast route requests for new route discoveries which are replied back by the destination or intermediate nodes if they have recently used a route to the destination. Nodes keep minimal routing table size with next hop entry for each destination. Sequence number ensures loop free routing. It is a good choice especially for event driven or periodic data driven WSN applications like smart metering. In this paper, a new ad hoc routing protocol (ETL-AODV) is proposed to achieve reliability, energy efficiency, and self-healing requirements of WSN based smart metering deployment. We have proposed modified routing criterion of AODV by utilizing multiple routing metrics approach and have presented simulation studies for performance analysis of our protocol.

The organization of the paper is as follows. In Section 2, we present background study on ad hoc routing protocols for smart metering application. Issues associated with deploying ad hoc routing protocol for WSN based smart metering infrastructure are highlighted, and current work on improving routing criterion of AODV is discussed. The proposed ad hoc routing protocol is presented in Section 3, and its performance, under different case studies, is presented in Section 4.

2. Background

Smart meters are provided with wireless interfaces to form mesh network. Due to inherent features such as the distributed nature of the network hosts along with redundancy and lack of single point of failure, the ad hoc networks exhibit robustness which is highly desirable for low cost commercial application. ad hoc networking represents a peer-to-peer multihop communication architecture. For each sink, there is a mesh network of metering nodes communicating with it in self-organized autonomous fashion.

Ad hoc routing algorithms dynamically determine the best path towards the destination based on the network status and in case of node failure, routing tables of nodes are updated to route the packet along alternative routes. Ad hoc routing protocols are classified into three types depending upon their route formation techniques as explained below.

(i)

Proactive (table driven): every node maintains a table of routes to every other node in the network and requires frequent exchange of topology messages. Examples: destination sequence distance vector (DSDV) [4], optimized link state routing (OLSR) protocol [5].

(ii)

Reactive (on demand): routes are formed only when required. Examples: ad hoc on-demand distance vector (AODV), dynamic source routing (DSR) [6].

(iii)

Hybrid: it combines characteristics of reactive and proactive protocols. Example: zone routing protocol (ZRP) [7].

Ad hoc network deployment of smart meters is greatly influenced by the nature of ad hoc routing protocol used. Proactive protocols optimize routing delays at the expense of bandwidth and power consumption while reactive protocols are bandwidth and energy efficient at the expense of route discovery delays. Smart metering devices are small in size and are cost effective due to the enormous deployment volume. Since smart metering deployment could consist of large number of nodes, the routing table that each node would have to keep could be huge, and therefore proactive protocols are not suitable for these kinds of networks.

2.1. Current Research on Wireless Mesh Routing Protocols for Smart Metering Infrastructure

Geelen et al. [8] proposed wireless mesh communication protocol for smart metering based on time synchronization. Metering nodes initiate communication with the concentrator nodes in predefined allotted slots. Nodes try to directly communicate with the target nodes, and, in case of communication failure, controlled flooding search is utilized in AODV style.

Distributed autonomous depth first routing (DADR) [9] is proactive distance vector routing protocol concerned with minimizing control overhead due to changing link conditions and provides at the most K possible paths for each destination. This protocol comes with increase in CPU and memory overheads of intermediate nodes due to additional state in data forwarding phase.

A hybrid routing protocol (Hydro) [10] is a link state routing protocol for low-power and lossy networks. It utilizes directed acyclic graph to build multiple routes to border routers. Nodes periodically piggyback topology reports on frequent data traffic to border router which in turn have global view of the network. This source routing can be a large overhead for networks with large number of nodes such as in case of smart metering network which may require many hops to reach the destination.

RPL [11] is distance vector routing protocol, currently under development, aimed for low-power and lossy networks such as home automation and industrial applications. It establishes directed acyclic graph based topology for supporting multiple sinks. A modified RPL based on ETX link metric was proposed for meeting reliability and latency requirements in AMI networks.

Gharavi and Hu [12] modified Hybrid Wireless mesh routing protocol (HWMP) for IEEE 802.11s based wireless mesh network (WMN) which establishes multiple paths to multiple gateways using proactive routing while on demand routing is initiated on path failures. Backup buffer stores their packets which are forwarded through backup routes in case of link failures. Jung et al. [13] proposed link error metric for IEEE 802.11s based WMN for accommodating varying packet size in smart grid scenario and route fluctuation prevention algorithm to improve overall reliability of the mesh network.

Li and Zhang [14] presented multiconstrained QoS routing for smart grid. It is based on simple greedy algorithm based on two QoS requirements of delay and outage probability. Lichtensteiger et al. [15] proposed geographic routing based mesh system for smart metering infrastructure and concluded that improved performance is achieved given that coverage gaps are filled. Ullo et al. [16] evaluated Zigbee based wireless sensor network for smart metering infrastructure and observed that congestion and delay increase with number of nodes.

2.2. Motivation

The routing protocols previously mentioned are aimed at increasing reliability of the network [17]. These protocols are focused on utilizing single routing criteria and are not designed with multiple routing metrics. In addition, they do not consider energy consumption of the nodes. This behaviour may become harmful for battery powered smart meters specifically those that are close to the sink as they will be involved in most of the multihop transmissions leading to fast energy depletion and congestion. In addition to this, link quality should be considered as well since frequent route breaks occur in wireless sensor networks due to fading effects and signal interference. Routing protocol should be aware of residual energy, traffic load, and link quality between the nodes to overcome the aforementioned issues. Taking this into mind, a new ad hoc routing protocol is proposed with route formation improved in such a way that nodes embed their residual energy, traffic load, and link quality information in the route request packets, and destination is able to select the best available path to the source.

2.3. AODV Based Routing Algorithm

Our protocol is based on the AODV routing protocol since its features are very much suitable for the smart metering infrastructure. Due to its reactive nature, no topology messages exchange is required for communication along the links which reduces bandwidth utilization particularly for large number of metering node network. AODV is favourable for resource constrained nodes such as battery powered smart meters due to its small routing table size wherein only next hop neighbour entry is maintained for each destination. The most important advantage of AODV is its ability to heal itself in case of nodes or routes failures. These are some of the advantages that make AODV a suitable candidate for smart metering infrastructure. AODV creates considerable routing overhead with relatively higher delay in route discovery phase as compared to other routing techniques. As our network is composed of static nodes with relaxed latency requirements so this issue may be overlooked for the advantages we gain. As suggested by research work in [18], AODV needs modification for use in smart metering infrastructure. AODV is based on minimum hop counting algorithm. However, minimum hop count routing does not always yield the best path. These shortest path routing protocols degrade network performance due to traffic congestion, unstable links, and power depletion on nodes along minimum hop path.

2.4. Prior Work on Routing Criterion Optimization of AODV

Originally proposed AODV algorithm selects routes based on minimum hop count. Many variants of AODV have been studied and proposed by the researchers utilizing various routing metrics for enhancing AODV performance. Most of the authors improved AODV performance using a single routing criterion. For instance, research works in [19–22] consider the energy level parameter of the nodes (residual or drain rate) as routing criterion for prolonging network lifetime. Similarly, the works in [23–26] consider the traffic load (buffer occupancy levels) of the nodes during route formation phase for reducing congestion. The works in [27–31] utilize the strongest link paths (based on received signal strength or SNR) for establishing stable routes.

The limitation of single routing metric for path selection has a side effect of overloading and depleting the resources along the selected path. Our work is concerned with utilizing multiple metrics for enhanced performance. Few variants of AODV consider multiple routing criterions. ETR-AODV [32] considers nodes residual energy and traffic load during route selection. R-AODV [33] proposed the idea of higher and lower level thresholds of battery level and signal strength. The thresholds decide when backup route should be searched and utilized. EM-AODV [34] is composed of composite metric of signal strength, battery power, and bandwidth. MMRP and its two variants (MMRP-I, MMRP-A) [35] are based on AODV and consider weighted contribution of hop count, traffic load, and residual energy of nodes for route selection.

3. Protocol Description

Our proposed protocol (ETL-AODV) is described as follows.

3.1. Routing Metrics

AODV was basically designed for mobile ad hoc networks (MANETs) and as such it only considers hop count during route discovery for finding shortest paths to the destination. This criterion needs to be modified for smart metering infrastructure since smart meter nodes are resource deficient and we need to consider energy consumptions of the nodes, traffic loads, and link quality during route selection for improving the performance.

3.1.1. Energy Level

Energy is a valuable resource for battery powered smart meters since their life directly depends upon how much energy is available to them. The energy level of the nodes involved in most of multihop transmission (like the ones closer to the sink) will be depleted quickly, and so the protocol should consider energy states of the nodes during route formation for forming the routes with highest energy nodes. Our protocol keeps track of the energy state of the nodes on the basis of residual energy of the node. Energy of the node is scaled on (0-1) scale using

\begin{matrix} E = \frac{E_{r}}{E_{\max}}, \end{matrix}

(1)

where $E_{r}$ is the remaining energy and $E_{\max}$ is the maximum energy available to the node. The energy value will vary from 1 to 0 with 1 corresponding to full energy level and 0 for all energy depleted (dead node).

3.1.2. Traffic Load

Traffic congestion is experienced by the nodes when incoming traffic is much higher than outgoing traffic. Smart metering nodes buffer the incoming packets in finite size queue and start dropping any new incoming packets when queue is full. Nodes which serve maximum number of nodes in multihop transmission (like the ones close to sink) are likely to experience maximum traffic load leading to traffic congestion. Traffic load of the nodes should be counted in during route selection for minimizing the congestion.

Our protocol measures traffic load of the node by measuring the interval between two received data packets and is normalized on (0-1) scale with 1 denoting minimum traffic load. Interval is updated using exponential smoothing function to avoid abrupt traffic jitters as shown in the following equation:

\begin{matrix} T_{F} = (1 - β) \times intv l_{old} + β \times intv l_{new}, \end{matrix}

(2)

where $intv l_{old}$ and $intv l_{new}$ are old and new intervals, respectively and β is adjustable parameter between 0 and 1. The larger the β value, the more sensitive to the new value. We have chosen β as 0.2 in our simulations.

3.1.3. Link Quality

Smart metering nodes communicate in outdoor environment, and active route breaks occur even on stationary nodes due to the effect of shadowing and consequently cause the network degradation. Shortest available path is not always the best path available since routing over short path with weak RF link quality leads to increased packet loss and retransmissions. Consequently, our protocol measures signal strength of the nodes while choosing candidates for route formation. Link quality ( $L_{Q}$ ) is scaled on (0-1) scale using the following equation:

\begin{matrix} L_{Q} = \frac{S_{P}}{S_{\max}}, \end{matrix}

(3)

where $S_{P}$ is the signal strength of the packets received and $S_{\max}$ is the maximum signal strength available. In this way, Link Quality parameter varies from 1 to 0 with 1 denoting best link available.

3.1.4. Multimetric Combination

Routing based on multiple criteria is a classification problem, and it complies to the additive combination rule [36], and we can combine multiple criteria (energy level, traffic load, link quality, and hop count) into a single criterion to get improved performance. Most of the multimetric routing protocols use the weighted sum approach to combine multiple metrics over an available path. In our protocol, ETL value is assigned to each node using the following general linear equation:

\begin{matrix} ET L_{node} = \sum_{i = 1}^{n} α_{i} m_{i}, \end{matrix}

(4)

where ETL is the value assigned to the node. “i” runs for all “n” number of metrics, and α is the weight assigned to metric “m”. Here in our protocol, we combine three metrics (energy level, traffic load, and link quality) with equal weights to get ETL value of a node as

\begin{matrix} ET L_{node} = W_{E} * E + W_{T} * T_{F} + W_{L} * L_{Q} . \end{matrix}

(5)

Here, we can assign priorities to different metrics depending upon application. Note that cumulative sum of the weights ( $W_{E}$ , $W_{T}$ , $W_{L}$ ) should always be equal to 1 so that ETL value obtained will be in range of (0-1). In this way, by selecting nodes with maximum ETL values for route formation, we can maximize the network performance.

3.2. Route Request Procedure

In AODV, whenever the source needs to find a route to new destination, it broadcasts a special route discovery packet (route Request packet—RREQ). Each node on receiving RREQ packet further propagates it until it reaches the destination. Nodes only reply back to RREQ packet if they know a route to destination or they are the destination. We have appended a new ETL field in RREQ packet to carry ETL information as shown in Figure 1.

Figure 1

RREQ packet format of ETL-AODV with ETL field appended.

Source node broadcasts RREQ packet with ETL value initialized to 1. Each intermediate node on receiving the RREQ packet appends their ETL value inside RREQ's ETL field by multiplying with the value contained in the packet. As a result, $ET L_{node}$ value of all the nodes along the path is multiplied to obtain the $ET L_{path}$ value for the path as follows:

\begin{matrix} ET L_{path} = \prod_{i = 1}^{k} ET L_{node}; \end{matrix}

(6)

here “i” runs for all “k” number of nodes in a specific path, and by taking the product we get values in range (0, 1). This is shown in Figure 2.

Figure 2

RREQ reception at intermediate node.

3.3. Route Selection by the Destination Node

In AODV, destination node on receiving the first route request replies back and discards any further route requests because of its shortest path formation behaviour. In our protocol, destination node on receiving first route request (RREQ) waits for a small amount of time $Δ t$ and then replies back for a path with maximum $ET L_{final}$ value as shown in Figures 3 and 4. $ET L_{final}$ is evaluated to cater for the hop count in route selection and is evaluated using the following relation:

\begin{matrix} ET L_{final} = k_{1} * ET L_{path} + k_{2} * HF, \end{matrix}

(7)

where $k_{1}$ and $k_{2}$ are weights with condition ( $k_{1} + k_{2} = 1$ ) and $HF$ is the hop factor calculated as

\begin{matrix} HF = \frac{H_{\max} - H_{count}}{H_{\max}}, \end{matrix}

(8)

where $H_{count}$ is the present hop count and $H_{\max}$ is the maximum hop count permissible by the protocol. Hop factor will have the value in range (0, 1) with direct links having HF as 1 and will keep on decreasing with the increase in the number of intermediate nodes. These weights can be varied to change importance of $ET L_{path}$ and the hop count during route selection. Here, we have initialized values of each one of them to 0.5. It is important to note that any other alternative function can also be used without impacting generality of the proposed solution.

Figure 3

RREQ selection by destination node.

Figure 4

Route reply generation by the destination node.

3.4. Low Energy Alert

Energy is the most important commodity for battery operated wireless networks, and therefore protocol design should be energy efficient. AODV has a drawback that a route once formed is continued to be used until transmission is completed or route breaks. Neglecting energy level of nodes in active routes may result in usage of all of their available energy. Our protocol tackles this problem by incorporating alert behaviour in the nodes. Our protocol keeps an eye on energy level of the node and sends an alert to the source node in case the battery level drops below a threshold value (Figure 5). Source node will discover some other route (if available) and saving the low battery node from becoming dead. Once the energy level of the node is improved, it can be included in future route formations.

Figure 5

Route change request at low energy.

3.5. MAC Level Acknowledgement

AODV utilizes periodic exchange of HELLO packets between neighbouring nodes for reliability purposes at the cost of increased routing overhead. Our protocol, instead of using HELLO packets, utilizes MAC level acknowledgement which leads to improved reliability as well as reduced routing overhead.

3.6. Route Reply by Destination Node Only

In our protocol, intermediate nodes are not allowed to reply back to RREQs even if they have route to destination, and they continue to propagate RREQ packet along the path until it reaches the destination node. This is to ensure that the highest performance route is selected by the destination node since it has the global view of all the nodes to make a wise decision.

4. Simulation Results

We have implemented our protocol in network simulator (NS-2). The smart meter nodes were configured with parameters derived from the specifications of a real smart meter [37]. Network simulation parameters are given in Table 1.

Table 1

Network simulation parameters.

Simulation parameters
Routing protocol	ETL-AODV, AODV
MAC layer/PHY layer	802.11
Channel type	Wireless channel
Propagation model	Two ray ground, shadowing
Traffic type	Constant bit rate (CBR)
CBR packet size	100 Bytes
CBR packet interval	1 sec, 60 sec
Route reply timer	0.1 sec
Low battery threshold	0.5
Antenna model	Omni Antenna

The performance metrics used are as follows.

(i)

Packet Delivery Ratio (PDR). It is calculated as the ratio of number of received data packets to the number of sent data packets. It measures how much of the sent data made up to the receiver thus representing reliability of the protocol. It is desirable to have maximum packet delivery ratio.

(ii)

Normalized Routing Load (NRL). It represents number of routing control packets exchanged in the network. It is calculated as ratio of number of received routing packets to the number of received data packets. It is desirable to have minimum routing overhead for the efficient operation of the network.

(iii)

Average Energy Consumption. This metric indicates energy consumed in the nodes of the network. Battery powered smart meters require energy efficient protocol operation for prolonging network lifetime.

(iv)

Average End-to-End Delay. This metric indicates latency in the communication network. It is calculated as the ratio of total time taken by all the packets to reach the destination to the total number of packets. Protocol should have minimum average delay for prompt data transfer. Although metering application is less sensitive to delays, however, outages notifications require to be transmitted without any delay.

We conducted three simulation based case studies for analyzing the performance of our proposed protocol. First simulation experiment was designed to mark performance based on individual metrics by setting desired metric's coefficient to 1 and others to 0. The second and third considered equal contribution of all metrics.

4.1. Case Study I

Three scenarios were simulated to analyze the effect of individual metrics (energy, traffic load, and link quality).

4.1.1. Scenario I

In Scenario I (Figure 6), we measured performance of the protocol based on energy criterion by setting ETL coefficients in (5) as (1, 0, 0). Initial energy of all the nodes is set to 10 joules except node 1 with 6 joules. When Node 4 wants to communicate with node 0, two routes can be formed (4-1-0) and (4-3-2-0). The AODV chose the shortest path thus draining energy deficient node 1 and consequently limiting network lifetime to 4500 sec and dropping PDR below 100%. In ETL-AODV, node 0 received two RREQs one via node 1 with $ET L_{Final} = 0.73$ and the other via node 2 with $ET L_{Final} = 0.95$ . Consequently, ETL-AODV chose high residual energy path (4-3-2-0) thus keeping PDR 100% and extending network lifetime (time interval from beginning till the death of first node) to 5500 sec. At time 2546.0028, energy of node 3 fell below threshold (0.50), and so it sent route Error to upstream node without breaking the connection and consequently on second route discovery, path (4-1-0) was formed. Results for PDR and energy consumption of nodes 1 and 2 are given in Figures 7, 8, and 9.

Figure 6

Scenario I.

Figure 7

PDR in Scenario I.

Figure 8

AODV energy consumption in Scenario I.

Figure 9

ETL-AODV energy consumption in Scenario I.

4.1.2. Scenario II

In Scenario II (Figure 10), we set the ETL coefficients in (5) as (0, 1, 0) for considering the effect of only the traffic load in routing decisions. Node 1 and node 4 are source meters with node 0 as sink node. Node 1 starts sending Data to sink at the start of simulation with 0.1 sec packet intervals depicting heavy load. We switch on node 4 a bit later at time 80 sec to study which route will be chosen by the protocols. Node 4 has two disjoint paths towards node 0 (4-1-0 and 4-3-2-0). The former path although the shortest path will increase loading on node 1 as it is also forwarding data to the sink leading to traffic congestion, fast energy depletions, and packet drops. AODV chose this path and consequently PDR dropped to 98.87%. In case of ETL-AODV, node 0 received two RREQs of node 4, one via node 1 with ${ETL}_{Final} = 0.47$ and the other via node 2 with ${ETL}_{Final} = 0.95$ . Consequently, ETL-AODV chose low traffic path (4-3-2-0) thus retaining PDR to 100%. The result of PDR is depicted in Figure 11. Similarly, ETL-AODV exhibited lower NRL and average end-to-end delay as compared to AODV (Figures 12 and 13).

Figure 10

Scenario II.

Figure 11

PDR in Scenario II.

Figure 12

NRL in Scenario II.

Figure 13

Average end-to-end delay in Scenario II.

4.1.3. Scenario III

In Scenario III (Figure 14), we studied effect of choosing link quality in routing decision by setting ETL coefficients in (5) as $(0,0, 1)$ . Furthermore, we chose shadowing propagation model to simulate an outdoor “shadowed urban area” with the parameters set as path loss exponent = 2.7 and standard deviation = 4. Node 4 is the source meter communicating with the sink node 0. There are two disjoint paths from node 2 to node 0 (4-1-0 and 4-3-2-0). The former one is the shortest but node 1 is deliberately placed further away to degrade Link Quality. As expected, AODV chose the weak link quality path. In case of ETL-ADOV, node 0 received two RREQs of node 4, one via node 1 with ${ETL}_{Final} = 0.47$ and the other via node 2 with ${ETL}_{Final} = 0.50$ . Consequently, ETL-AODV chose high link quality path (4-3-2-0) thus retaining PDR to 100%. Average end-to-end delay and normalized routing load were very small as compared to that of AODV. The results for PDR, NRL and average end-to-end delay are shown in Figures 15, 16 and 17.

Figure 14

Scenario III.

Figure 15

PDR in Scenario III.

Figure 16

NRL in Scenario III.

Figure 17

Average end-to-end delay in Scenario III.

4.2. Case Study II

In Case Study II, we considered all the three parameters of ETL metric with equal contribution by setting ETL coefficients in (5) as (1/3, 1/3, 1/3). We simulated topologies based on real geographic suburban region with location of houses as wireless metering nodes as shown in Figure 18 and star denoting location of sink node.

Figure 18

Geographical area with houses as nodes position.

We increased the number of houses from 25 to 125 in regular grid structure and evaluated the performance of our proposed protocol. The sending interval was set as 60 seconds, and all of the metering nodes were configured to start sending data to sink at time chosen from uniform random distribution (0–60) sec. We chose shadowing propagation model to simulate an outdoor “shadowed urban area” with the parameters set as path loss exponent = 2.7 and standard deviation = 4. All the simulations were run for 1 hour. We simulated each topology 10 times with different seed value and averaged the result. The results for PDR, NRL, average energy consumption and average end-to-end delay are given below.

4.2.1. Packet Delivery Ratio (PDR)

The results of PDR for two protocols simulated are shown in Figure 19.

Figure 19

PDR versus number of smart meters.

A general decreasing trend is observed for both protocols with increase in number of metering nodes. However, for each set of nodes, ETL-AODV outperforms AODV protocol. This is due to the fact that ETL-AODV considers energy, traffic load, and link quality during route formation so more stable routes are formed thus enhancing PDR. On the other hand, AODV forms only the shortest route which can have negative impact on PDR. ETL-AODV maintained 80% PDR for 125 number of homes in shadowed region as compared to AODV whose PDR declined to 69.8%.

4.2.2. Normalized Routing Load (NRL)

The results for NRL are shown in Figure 20. As predicted, ETL-AODV has much lower routing overhead as compared to AODV. This is due to formation of stable routes by considering all three submetrics leading to reduced exchange of control packets.

Figure 20

NRL versus number of smart meters.

4.2.3. Average Energy Consumption

The results for average energy consumption of nodes are shown in Figure 21. ETL-AODV has much better energy consumption as compared to AODV due to inclusion of energy metric in routing decisions. This metric is much useful for battery powered smart meters such as waters and gas meters for prolonging network and node lifetime.

Figure 21

Average energy consumption versus number of smart meters.

4.2.4. Average End-to-End Delay

The results for average end-to-end delay are shown in Figure 22. Both protocols show increasing trend with increase in number of houses. ETL-AODV exhibits a bit higher delay due to additional route selection phase in destination node. However, as time constraint is not an issue in smart metering applications so this can be compromised for improved PDR, NRL, and energy consumption of nodes.

Figure 22

Average end-to-end delay versus number of smart meters.

4.3. Case Study III

In Case Study III, we analyzed the effect of intermeter distances on performance metrics for both protocols. We chose 50 homes topology and simulated the network for one hour with same simulation parameters as of case study II. The results for PDR, NRL, average energy consumption and average end-to-end delay are given below.

4.3.1. Packet Delivery Ratio (PDR)

The results for PDR are shown in Figure 23. Both protocols follow declining trend with the increase in intermeter distances. This is due to the fact that when nodes are close to one another in ad hoc network, variety of routes are available thus enhancing PDR. As interdistance between nodes starts to increase, PDR is severely affected due to shadowing effects. ETL-AODV outperforms AODV for all three intermeter distances due to formation of stable paths.

Figure 23

PDR versus intermeter distance.

4.3.2. Normalized Routing Load (NRL)

The results for NRL for ETL-AODV and AODV are plotted in Figure 24. NRL gradually increases with the increase in intermeter distances. However, significant improvement is observed in case of ETL-AODV due to improved route discovery leading to less control overhead. Formation of stable optimum paths leads to less route breakages and ultimately low NRL.

Figure 24

NRL versus intermeter Distance.

4.3.3. Average Energy Consumption

The results for average energy consumption for both protocols are plotted in Figure 25. As observed, ETL-AODV has much lower average energy consumption as compared to AODV which increases with increasing intermeter distances. Unstable routes are prone to frequent breakages due to shadowing effects leading to high overhead route discoveries which increase energy consumption of nodes. This is harmful for battery operated smart meters such as gas and water smart meters.

Figure 25

Average energy consumption versus intermeter distance.

4.3.4. Average End-to-End Delay

The results for average end-to-end delay are plotted in Figure 26. ETL-AODV has higher average end-to-end delay as compared to AODV for 30 m intermeter spacing due to additional route selection waiting time at sink node. However, latency in AODV is higher as compared to ETL-AODV at 50 m and 70 m intermeter spacing. This is due to the fact that at large distance, links are prone to more breakages due to link fluctuations results from shadowing effects which severely degrades AODV performance.

Figure 26

Average end-to-end delay versus intermeter distance.

5. Conclusion and Future Work

In this paper, we have proposed a reliable WSN ad hoc routing protocol (ETL-AODV) for smart metering infrastructure. The proposed protocol is based on AODV designed for low cost and resource constrained smart metering sensor nodes. Enhanced reliability is achieved by considering energy level, traffic load, and link quality of the nodes to establish appropriate route from smart meter to the sink. Our simulations studies indicate that the use of the aforementioned parameters for route selection results in improved packet delivery ratio and reduced energy consumption of metering sensor nodes.

For future work, we wish to implement our proposed protocol on hardware testbed to consolidate simulation results.

Footnotes

Acknowledgment

The authors are grateful to Universiti Teknologi Petronas, Malaysia, for sponsoring this research work through their Graduate Assistantship Scheme.

References

Yeung

Jung

Improving electric reliability with smart meters

Whitepaper, Silver Spring Networks

Farooq

Jung

L. T.

Health, link quality and reputation aware routing protocol (HLR-AODV) for Wireless Sensor Network in Smart Power Grid

Proceedings of the International Conference on Computer & Information Science (ICCIS '12)

June 2012

664 669

Perkins

Royer

E. B.

Das

AdHoc on-demand distance vector (AODV) routing

RFC 2003 3561

IETF Network Working Group

Perkins

C. E.

Watson

T. J.

Highly dynamic destination sequenced distance vector routing (DSDV) for mobile computers

Proceeding of the ACM SIGCOMM Conference On Communications Architectures

1994

234 244

Clausen

Jacquet

The optimized link state routing protocol (OLSR)

IETF RFC 2003 3626

Johnson

D. B.

Maltz

D. A.

Y. C.

The Dynamic Source Routing Protocol For Mobile Ad Hoc Networks (DSR)

Internet Draft, Draft-Ietf-Manet-Dsr-09.Txt, 2003

Haas

Z. J.

Pearlman

M. R.

Samar

The Zone Routing Protocol (ZRP) for Ad Hoc Networks

draft-ietfmanetzone -zrp- 04. txt, 2002

Geelen

Van Kempen

Van Hoogstraten

Liotta

A wireless mesh communication protocol for smart-metering

Proceedings of the International Conference on Computing, Networking and Communications (ICNC '12)

February 2012

343 349

2-s2.0-84859900786

10.1109/ICCNC.2012.6167440

Iwao

Yamada

Yura

Nakaya

Cardenas

Lee

Masuoka

Dynamic data forwarding in wireless mesh networks

Proceedings of the 1st IEEE International Conference on Smart Grid Communications (SmartGridComm '10)

October 2010

385 390

10.

Dawson-Haggerty

Tavakoli

Culler

Hydro A hybrid routing protocol for low-power and lossy networks

Proceedings of the 1st IEEE International Conference on Smart Grid Communications (SmartGridComm '10)

October 2010

268 273

11.

Winter

Thubert

Brandt

Clausen

Hui

Kelsey

Levis

Pister

Struik

Vasseur

Rpl: Ipv6 routing protocol for low power and lossy networks

2011

12.

Gharavi

Multigate communication network for smart grid

Proceedings of the IEEE 2011 99 6 1028 1045

2-s2.0-79956370485

10.1109/JPROC.2011.2123851

13.

Jung

J. S.

Lim

K.-W.

Kim

J.-B.

Y.-B.

Kim

Lee

S. Y.

Improving IEEE 802.11 s wireless mesh networks for reliable routing in the smart grid infrastructure

Proceedings of the IEEE International Conference on Communications Workshops (ICC '11 Workshops)

June 2011

1 5

2-s2.0-80051968863

10.1109/iccw.2011.5963578

14.

Zhang

QoS routing in smart grid

Proceedings of the 53rd IEEE Global Communications Conference (GLOBECOM '10)

December 2010

1 6

2-s2.0-79551627452

10.1109/GLOCOM.2010.5683884

15.

Lichtensteiger

Bjelajac

Muller

Wietfeld

RF mesh systems for smart metering: system architecture and performance

Proceedings of the 1st IEEE International Conference on Smart Grid Communications (SmartGridComm '10)

October 2010

379 384

16.

Ullo

Vaccaro

Velotto

The role of pervasive and cooperative sensor networks in smart grids communication

Proceedings of the 15th IEEE Mediterranean Electrotechnical Conference (MELECON '10)

April 2010

443 447

2-s2.0-77954309031

10.1109/MELCON.2010.5476236

17.

Saputro

Akkaya

Uludag

A survey of routing protocols for smart grid communications

Computer Networks 2012 56 11 2742 2771

2-s2.0-84860591328

10.1016/j.comnet.2012.03.027

18.

Farooq

Jung

L. T.

Performance analysis of AODV routing protocol for wireless sensor network based smart metering

Proceedings of the IOP Conference Series: Earth and Environmental Science

2013

19.

Veerayya

Sharma

Karandikar

SQ-AODV: A novel energy-aware stability-based routing protocol for enhanced QoS in wireless ad-hoc networks

Proceedings of the IEEE Military Communications Conference (MILCOM '08)

November 2008

1 7

2-s2.0-62349101846

10.1109/MILCOM.2008.4753608

20.

Tie

T. H.

Tan

C. E.

Lau

S. P.

Maximum energy level ad hoc distance vector scheme for energy efficient ad hoc networks routing

Proceedings of the IEEE 9th Malaysia International Conference on Communications with a Special Workshop on Digital TV Contents (MICC '09)

December 2009

423 428

2-s2.0-77952174619

10.1109/MICC.2009.5431544

21.

Chen

Energy efficient routing protocol for Ad hoc networks

Proceedings of the International Conference on Computer Design And Appliations

2010

320 323

22.

Dhurandher

S. K.

Obaidat

M. S.

Misra

Bansal

Singh

Punia

An energy-efficient on-demand routing algorithm for mobile ad-hoc networks

Proceedings of the 15th IEEE International Conference on Electronics, Circuits and Systems (ICECS '08)

September 2008

958 961

2-s2.0-57849143946

10.1109/ICECS.2008.4675014

23.

Xia

Liu

Chang

Sun

An improved AODV routing protocol based on the congestion control and routing repair mechanism

Proceedings of the WRI International Conference on Communications and Mobile Computing (CMC '09)

January 2009

259 262

2-s2.0-64749102993

10.1109/CMC.2009.307

24.

Yuan

Chen

Jia

An adaptive load-balancing approach for ad hoc networks

Proceedings of the 2005 International Conference on Wireless Communications, Networking and Mobile Computing (WCNM '05)

September 2005

743 746

2-s2.0-33646937670

25.

Ahmad

Ur Rehman

Efficient AODV routing based on traffic load and mobility of node in MANET

Proceedings of the 6th International Conference on Emerging Technologies (ICET '10)

October 2010

370 375

2-s2.0-78650334696

10.1109/ICET.2010.5638459

26.

Shi

Zong

AODV protocol optimization based on cross-layer design of WMN

Proceedings of the 7th International Conference on Wireless Communications, Networking and Mobile Computing (WiCOM '11)

September 2011

1 4

2-s2.0-80054924280

10.1109/wicom.2011.6040282

27.

Anis

B. A.

Nguira

Habib

Wireless routing protocol based on auto - learning algorithm

Proceedings of the 3rd International Conference on Digital Society (ICDS '09)

February 2009

118 122

2-s2.0-63749108567

10.1109/ICDS.2009.64

28.

Chang

R.-S.

Leu

S.-J.

Long-lived path routing with received signal strength for ad hoc networks

Proceedings of the 1st International Symposium on Wireless Pervasive Computing

January 2006

tha

2-s2.0-33750418800

29.

Wang

S.-Y.

Liu

J.-Y.

Huang

C.-C.

Kao

M.-Y.

Y.-H.

Signal strength-based routing protocol for mobile ad hoc networks

Proceedings of the 19th International Conference on Advanced Information Networking and Applications (AINA '05)

March 2005

17 20

2-s2.0-33744473820

10.1109/AINA.2005.311

30.

Park

Moh

Chung

A multipath AODV routing protocol in mobile Ad hoc networks with sinr-based route selection

Proceedings of the IEEE International Symposium on Wireless Communication Systems (ISWCS '08)

October 2008

682 686

2-s2.0-62449275771

10.1109/ISWCS.2008.4726143

31.

Moh

Kang

Chung

Wang

Link quality aware robust routing for mobile multihop Ad Hoc networks

Mobile Ad-Hoc Networks: Protocol Design 2011

Rijeka, Croatia

InTech

32.

Kumar

Prakash

Kumar Singh

An energy and traffic aware routing approach as an extension of AODV

International Journal of Computer Applications 2011 27 7 6 10

33.

Kumar

Rani

Kumar

Reliable-AODV

Proceedings of the 2nd International Conference on Computer and Communication Technology (ICCCT '11)

September 2011

568 573

2-s2.0-82955228081

10.1109/ICCCT.2011.6075108

34.

Thanthry

Kaki

S. R.

Pendse

EM-AODV: Metric based enhancement to AODV routing protocol

Proceedings of the IEEE 64th Vehicular Technology Conference (VTC '06 Fall)

September 2006

2594 2598

2-s2.0-34548843341

10.1109/VTCF.2006.534

35.

Cao

Sharif

Wang

Dahlberg

Adaptive multiple metrics routing protocols for heterogeneous multi-hop wireless networks

Proceedings of the 5th IEEE Consumer Communications and Networking Conference (CCNC '08)

January 2008

13 17

2-s2.0-51949111692

10.1109/ccnc08.2007.9

36.

Wang

Crowcroft

Bandwidth-delay based routing algorithms

Proceedings of the 1995 IEEE Global Telecommunications Conference. Part 2 (of 3)

November 1995

2129 2133

2-s2.0-0029508929

37.

Silverspring Networks Communications Module for Electricity Meters (datasheet)

http://www.silverspringnet.com/pdfs/SilverSpring-Datasheet-Communications-Modules.pdf