Novel energy aware routing protocol for multievent wireless sensor network

In this paper, we propose a novel solution combining an event driven routing protocol, dynamic delivering scheme, and energy aware to support QoS requirements for three event types in multiple event WSN. Simulation results show that, the proposed solution significantly reduces packet loss rate for high reliability requirement events and extends the network lifetime of multievent WSN.

NOVEL ENERGY AWARE ROUTING PROTOCOL FOR

MULTIEVENT WIRELESS SENSOR NETWORK

Nguyen Thi Thu Hang*, Nguyen Chien Trinh, Nguyen Tien Ban

Abstract: Multievent wireless sensor networks (WSN) such as smart buildings,

intelligent environmental monitoring systems require different QoS (Quality of Service) provision based on various event types These networks contain large numbers of sensor nodes but they have a very limited power and processing capability, so efficient consumption is one of vital requirements for most WSNs Most of research papers in this area have dealt with one or two of QoS requirements or with a limited number of event types and event sources For this reason, in this paper, we propose a novel solution combining an event driven routing protocol, dynamic delivering scheme, and energy aware to support QoS requirements for three event types in multiple event WSN Simulation results show that, the proposed solution significantly reduces packet loss rate for high reliability requirement events and extends the network lifetime of multievent WSN Moreover,

in case of high traffic load condition, sharing load over multiple paths would decrease latency for the urgent events in the multiple events network

Keywords: Energy aware routing, Dynamic routing, Delivering scheme, Multievent, Wireless sensor network

1 INTRODUCTION

In some wireless sensor networks (WSN), there are different types of events based on their important levels Important events can be considered as abnormal situations Poisonous gas or liquid detection in chemical industry, fires in forest fire alarm systems are such kinds of events [1, 2] If the leakage occurs or wildfire happens, the monitor system must know it immediately Sometimes there may be several leaking points or wildfires, so there are multiple events appear in the network Then, it is more urgent to locate all of them It needs not only to locate the leaking points or hot spots but also to tell the volumes of leak or the burn areas Other environmental parameters, such as temperature, pressure, humidity, light intensity, and so on, can also be monitored and considered as normal events

With WSNs for smart buildings, intelligent environmental monitoring, and industrial process [1-7], multiple events with different levels of importance may happen in the networks Take an example of forest fire alarm system, forest fire risk usually occurs during and after winters with little rain, after long periods of dry weather and during summer heat waves, and especially if such conditions coincide with strong winds The forest fire risk indicates the probability of a forest fire occurring For the forest fire alarm system, there are five danger levels of forest fire: level 5 (very high): fires can start

at any time, the sensor data must be transmitted quickly to the base station; level 4 (high) and level 3 (considerable): the sensor data should reach the sink with high reliability because it could indicate a possibility of forest fire, level 2 (moderate) and level 1 (low or none): the data is not too serious, so it can be transmitted without specific requirement of low latency or high reliability [8] Fire spots can appear in many different areas making various events with different levels of QoS requirements such as latency and reliability For most WSNs, energy efficient consumption is one of crucial requirements because sensors have limited power and processing capability [3] The wireless sensor node can only be equipped with a limited power source and in some application scenarios, replacement of power resources might be impossible Sensor node lifetime, therefore,

shows a strong dependence on battery lifetime So, many researchers have been focusing

on the design of power-aware protocols and algorithms for sensor networks [9-11]

To meet these requirements of QoS and energy efficiency, there are three essential approaches as follows

First, for providing different levels of reliability requirements, there have been many techniques that many researchers are interested in, in which routing is one of the most important techniques There have been many research papers on single path routing and multipath routing protocols [12-17] Although the work of finding a single path is simple with low computational complexity and minimum resource utilization [12], [13], it could react slowly with the rapid change in the network topology (node or link failure) and can not support reliability as required by limited capacity of a single path [c11] So, many multipath routing protocols have been researched and developed to overcome the disadvantages of the single path routing protocols [15-17] In the case of many event types appear in the network which have different requirement of reliability, the dynamic routing scheme which combines single path for normal event type and multipath for high reliability requirement event type can be applied [14, 18]

Second, splitting traffic over multiple paths could support the bandwidth requirements

of different applications and reduce the probability of network congestion, then reduce network latency [19]

Third, a lot of energy-efficient routing protocols have been proposed, they have been categorized and described in [10], [20-24], all of the protocols aimed at energy efficient consumption and expanding the network life time Besides, the technique of transmitting multiple copies of data packets over multiple paths in [14] will increases delivery reliability but the energy consumption would be much more times, that is a trade-off between energy and reliability So, applying energy-aware with event driven routing would even be more necessary to increase the energy efficiency in such multiple event WSNs

To the best of our knowledge, all of research papers in this area have dealt with one or more requirements, and dealt with limited events and types of events There has been one research that raised the issue of challenges between a single-event wireless sensor network and multi-event wireless sensor network [23], but in the probable situation of concurrent events in the network, the research showed that it was unable to benefit effectively for data transmission over multiple paths than over single path, it provided shorter life time This

is the first work that uses energy aware dynamic routing and packet delivering schemes to support the multi QoS requirements for multiple event type WSN

In this paper, we proposes a combined solution for QoS provision, named EARPM (Energy Aware Routing Protocol for Multievent Wireless Sensor Network) for multievent wireless sensor network: to choose dynamic routing protocol and packet delivering scheme in WSN based on residual energy of nodes and different event types Our contributions in this paper are as follows:

1 We propose a combination of dynamic routing scheme of single and multipath, and different packet delivering schemes of copying or splitting packets based

on three event types to support the different reliability and latency requirements in multievent WSN

2 We also propose an energy-aware algorithm to dynamically discover energy efficient paths for delivering event packets

3 We implement our proposed routing and delivering schemes in OMNeT++ simulation to evaluate the adaptation of the network to the multiple event requirements and the efficiency of the energy aware scheme

The paper is organized as follows: Section 2 discusses the related work Section 3 describes our proposed solution Section 4 introduces our theory analyses The evaluation

of our protocol based on computer simulation is presented in Section 5 Finally the last section is the summarization and our future research work

2 RELATED WORK

Recently, there have been several research papers on multipath routing protocols and energy aware routing protocols to achieve various performance benefits

In ReInForM (Reliable Information Forwarding Using Multiple Paths [14]), the source sends multiple copies of the same data through multiple paths to the sink Each packet is assigned a priority level based on the content of the information it contains The source computes the number of paths (or equivalently, the number of copies of the packet to be sent) based on the importance of the information, local channel error and distance from the sink ReInForM does not distinguish between the actual source and an intermediate forwarding node Next hops are usually chosen among the nearest hops to the sink, otherwise they would be chosen randomly This helps in load balancing and avoids the nodes on the “better” path to be quickly energy depletion However, sending multiple copies of all packets would waste energy and the routing protocol has not considered the latency of the event The research has considered only single event source scenario, not multiple events

A low-interference energy-efficient multipath routing protocol (LIEMRO) has been designed for improving QoS in event-based WSN [23] This protocol has discovered multiple interference-minimized node disjoint paths between source node and sink node and included a load balancing algorithm to distribute source node's traffic over multiple paths based on the relative quality of each path The simulation shows that in high traffic load conditions, it can increase data reception rate, lengthen the network life time, and significantly reduce end-to-end latency compared with single path routing approach The research has raised the issue of challenges between a single-event wireless sensor network and multi-event wireless sensor network LIEMRO tries to construct node-disjoint paths for each detected event Nevertheless, paths with shared nodes are probable when two or more events occur in the network Therefore, the research has also evaluated LIEMRO in multiple event situations The simulated results show that in this situation, LIEMRO is unable to benefit effectively for data transmission over multiple paths than over single path, it provides shorter life time

In [18], a multipath routing protocol has been proposed in which the sink discovers paths based on path weight factor by using link efficiency, energy ratio, and hop distance The sink selects the number of paths among the available paths based upon the criticalness

of an event, and if the event is non-critical, then single path with highest path weight factor

is selected, otherwise multiple paths are selected for the reliable communication So this research has just differentiated two types of events

In [25] a distributed, scalable and localized multipath search protocol has been introduced to discover multiple node-disjoint paths between the sink and source nodes In this research, a load balancing algorithm is used to distribute the traffic over the multiple paths discovered, it allows the sink node to allocate traffic based on paths' cost, which depends on the energy levels and the hop distances of nodes along each paths The

proposed scheme has been compared to the directed diffusion [26], directed transmission, and the energy aware routing [9] protocols Simulation results show that it has higher node energy efficiency, low average delay But the research uses limited number of sink-source scenario, one sink with two or four sink-sources, two sinks with three sink-sources, and has not considered different packet types

From the above analyses, it can be seen that all of these research works have just dealt with only one or two event types which require QoS requirements of latency and/or reliability, some work has considered the energy efficiency but has not investigated the scenario of concurrent events, there has not been any research supported diversity QoS requirements for multievent WSN

Our proposal in this paper is to discovering energy-aware single and multiple paths, and use dynamic load delivering scheme which adapt to the three types of events, consequently it supports better performance for different event requirements of reliability, latency and energy efficiency for multievent WSN

3 PROPOSED SOLUTION

Based on the variety QoS requirements of multievent WSN and the benefits in getting high reliability and low latency of multipath routing protocols, we propose our novel energy aware dynamic routing protocol for multievent WSN

Our routing protocol is a renovation work from GPSR single path routing protocol [27] for event trigger routing WSN, so only greedy forwarding technique is applied when event appears in the network There are three dynamic changes have been done for the scheme First, based on the type of events, source node chooses single path for normal event type (named A, which does not require high reliability and low latency), multiple paths for the higher requirement event types (named B, which requires higher reliability, and C, which requires lower latency because of its urgency) Second, the delivering schemes are different from B and C: for B, data packets from source nodes should be copied and forwarded over two paths simultaneously while for C, packets should be split and sent over two paths Third, to avoid quickly depleted energy node on the shortest path, nodes in the network would choose the relay node(s) having residual energy more or equal to the average residual energy of all live and sink-nearer neighbors

We consider the average value of energy because time after time, the relay node will turn over among alive neighbors due to their residual energy levels have decreased by the time packets of an event passed by, so nodes will deplete their energy more slowly and equally Choosing the average value is better than choosing the highest residual energy value because the highest residual energy neighbor node might have the longer distance to the sink, so the energy consumption would be higher Furthermore, that highest residual energy neighbor node could be the good neighbor of other event node both in energy and distance to the sink in the multiple event network, so it should be chosen as the relay node

of the other node

Fig.1 shows a description of our dynamic routing schemes in multievent WSN Source nodes have to find the best neighbor(s) among the sink-nearer ones to deliver its sensed data packets and relay nodes have to find only one best neighbor There are five alive neighboring nodes (1, 2, 3, 5, 9) and one dead node (12) of the source in which only four nodes are alive sink-nearer (1, 2, 3, 9)

 For single path GPSR routing: there is one that is alive and nearest to the sink (node 3) So, source node would choose node 3 to be the best relay node on the routing path to the sink (Fig 1.a)

d max

2 4

5 6

d Source-BS

8

Source A

SINK

9

11

12

13

a) Single path GPSR routing

b) Multipath routing

c) EARPM routing

Figure 1 A description of the

combining of energy aware single and multiple path routing schemes

d max

2 4

5 6

d Source-BS

8

Source B/C

SINK

11

9 12 13

d max

2 4

5 6

d Source-BS

8

Source A/B/C

SINK

11

9

12

13

E >E >E >E =E

2 9

A,B,C B,C

3 1

average remained energy

 For multipath routing: the four alive

nodes can be chosen in priority order of

3, 2, 1, and 9 if only shortest distance

evaluation is used (Fig 1.b)

 For EARPM: we consider three criteria

in order of priority: (1) neighbor’s

residual energy, (2) distance from

neighbor node to the sink, and (3)

distance from source node to neighbor

node Then, at time, the residual energy

of node 2 is the highest and node 9 is

the second highest, the residual energy

of node 3 is equal to node 1, the

distances from neighbor nodes to sink

are in order 3, 2, 9, 1 as closer to the

sink, and the distances from source

node to its neighbors are in order of 1,

9, 2, 3 as nearer to the source Then,

the priority order of paths is 2, 9, 3,

and 1 Source A would choose 2 as the

delay node while source B and C

would choose 2 and 9 as the delay

nodes (Fig 1.c)

A Network Model

The WSN can be viewed as an undirected

graph G V E, where V represents the set

of vertices (sensor nodes and sink) and E

represents the set of edges We assume there

are N S sensor nodes randomly place in an area

2

node i and node j if the Euclidean distance

, 

node’s radio transmission radius dmax There

is a single monitoring node (sink) at the center

of sensing area, it knows its position and all

nodes’ position When sensor node detects an

event, it will send its data directly to the sink

if its distance to sink is less or equal to its

transmission range or indirectly over its

neighbors otherwise

B Energy Model

In our work, a simple radio model where

the radio dissipates E elecenergy per bit to run

the transmitter or receiver and amp energy

per bit for the transmit amplifier We also assume 2

d energy loss due to channel

transmission [28] So, the energy consumption to send a L-bit packet to next hop at a distance of d is:

2

2

E  E  L   L d (1)

C Proposed Routing Scheme

Fig 3 shows a brief description of our EARPM operation when node detects an event

or receives routing request from its neighbor node, then it has to select relay node(s) for delivering sensed data packets afterward

When sensor node detects an event, it will send routing requests to all of its alive neighbors, then all alive neighbors will send their routing requests toward all of their alive neighbors and so on After that, the source and all other related nodes will receive reply packets with the information of their neighbors’ residual energy to determine the node(s) which is/are eligible to be selected as next hop relay If a node’s residual energy E residual is less than E dead then it can not send reply REQ message, if the residual energy is less than

threshold

E then node can not send or forward data packets Only source node has to decide the number of paths for its sensed data based on the event type while all forwarding nodes have to choose only one best relay node

 If the distance to sink is equal or less than d max (the maximum transmission range of sensor), then node directly sends data to the sink

 If not, sensor node will have to find the best neighbors to deliver its data to the sink One or two best neighbors will be chosen based on three criteria: first, its/their residual energy (the neighbor’s residual energy is equal or larger than the average energy of all alive sink-nearer neighbors eResidualNeighbor[i]>= tempEAvg); second, among the neighbors that satisfy the first criteria, one or two neighbors that have shortest distance to the sink (as closer to the sink as possible) would be chosen; third, if there are neighbors that satisfy the previous two criteria, the order of best neighbors would depend on the distance from a neighbor to the source node (as closer as possible)

In this section, we analyze packet latency and make the probabilistic formulation of reliability for both single-path and multipath routing The results show that load-sharing

on multipath would reduce the queuing time of packets in congested situation, then reduce the packet latency in a simple way, and multi-path routing with redundant transmission is effective in increasing the reliability

Figure 2 Energy model – first order radio model

Transmit Electronics

Eelec*L

Tx Amplifier

amp*L*d 2

Eelec*L

Receive Electronics

d

L bit packet

L bit packet

Do not have to build routing table, sink can

be reached directly

N

d2SINK≤d max

Y

Y

Y

Send back a REP message with information of its residual energy

1 Send REQ messages to all of its alive

neighbors

2 Receive REP message(s) from the

neighbor(s)

3 Calculate the average residual energy of its

alive neighbor(s)

4 Maximum two best neighbors would be

selected as relaying node(s) based on alive

neighbors’ residual energy and distances

Node has enough energy to deliver data packet or not?

END

for (i=0;i<numOfneighbor;i++) {

if (eResidualNeighbor[i]>=eThreshold) {

tempE=eResidualNeighbor[i];

tempETotal= tempETotal+tempE;

numOfneighborLive++;

} } tempEAvg= tempETotal/numOfneighborLive;

N

BEGIN

Node detect event

Y

Node has enough energy to reply back or not?

Node receive REQ message (from other node)

remain threshold

E ≥ E

remain dead

E ≥ E

Figure 3 Description of relay selection operation

4 THEORETICAL ANALYSES

A Latency analysis

The total delay, denoted as d, experienced by a packet in a path of hop count h is the sum of the delays at the intermediate nodes, d j (where j1, 2, ,h ), and is given by

1

h j j



 (2)

Considering the propagation and processing delays as negligible, d j can be calculated

as follows

j trans MAC que

where d transis the transmission delay, d MAC is the medium access delay and d queis the queuing delay of a packet

In this paper, we concentrate into the queuing delay of a packet Queuing delay at any node depends on the queue service time, number of packets in queue, and the packet arrival pattern

Fig 4 shows the analysis of the queuing delay of packets We compare the queuing delay of packets over single and multiple paths using redundant transmission and load-sharing schemes From source nodes, there are three event type packets that would enter queues with the current queue length of *

Q packets over a maximum capacity of Q

packets

As we can see from Fig 4, for event type A and B packets, only N packets would be sent over one path, so the average queuing delay of packet type A and B is equal and can be approximately calculated as the delay of the middle packet  N / 2  For type C, it is less and proportional to the inversion of M - the number of paths, which can be calculated as

2

queA queB service

N

2

N

M

Figure 4 Occupation of queue for the three event types

If we denote the improvement of latency of C over A is l improvement, we can evaluate the improvement at one queue as follows

queB queC improvement

queB

N N

l



(6)

Let take Q* x N , then Eq.6 can be shortened as

e c

h hops

Reliability (1-e c ) (1-e c ) (1-e c j ) (1-e c )

Figure 6 Single path scenario

Figure 5 Latency comparison of multipath routing

using load sharing technique over single path routing with different numbers of paths and queue length

1 1

100%

improvement M l

x



 

(7)

From Eq 7, it is clear that splitting data packets over multiple paths would decrease the load placed on each link, thus

reducing the packet processing

time, the larger the number of

multipath (larger M), the better

the advantage of latency of C

over A, the larger the queue

(larger x), the lesser the

advantage of latency of C over

A Based on the fact that sensor

nodes have limited memory [3,

11], we can see that the sensor

queue capacity can not be so

large, so the value of l improvement

can have a great value

Fig 5 is a specific example

for the mathematical latency

comparison of multipath routing

using load sharing technique

over single path routing with

different numbers of paths and

queue length

B Reliability analysis

If the number of original packets sent by the source isN S, and the number of distinctive packets received by the sink isN R, the reliability, denoted as R, is RN r/N s Here the distinctive packet means that if sink receives multiplicative packets (the original data packet and the copy one), it considers those as one received packet

Reliability of Single-Path Routing

Consider a source and a sink which are h hops apart as in Fig 6 Let the per hop channel packet error rate (PER) at

th

j hop in the path across the entire

network be a variable c

j

e (where

c

j

0e 1, and it is proportional to

the distance), then the per-hop

reliability at th

j hop is (1 e c j)

The reliability of a path is a

multiplicative metric Thus, the probability that a packet is received

by the sink over a single path of h hops apart, p h , is

1 1

h c j j



Then single path packet error rate in this situation is

Figure 7 Multipath scenario.

e c 1,1

e c 2,j e c 2,h2

e c 2,1 e c 2,2

e c 1,2 e c 1,j

e c 1,h1

e c M,1

e c M,2 e c M,j

e c M,hM

Figure 8 Packet error rate evaluation based on the number

of hops, paths for per-hop channel error rate of 1%

 

1

h

j j



Thus, in a multi hop sensor network, where channel errors could be very high and a source could be far away from the sink, a nạve forwarding scheme will result in a high PER, so single path routing is not proficient of attaining good reliability

Reliability of Multipath Routing

Consider multiple paths from a

source to as in Fig 7 There are M

paths and the hop count of the i th

path is h i, the multipath packet

error rate in this situation is the

probability that all copy packets

would suffer error in all paths

The reliability of sending a

packet by copying it and send over

multiple paths can be calculated as

i h

where p h i i is

the probability of

success for the

t h

i path defined in

Eq 8 and c,

i j

e is the

probability that a

packet is dropped

at the th

j hop of the

th

i path

probability that at

least one copy of a

successfully

received by the

sink over M paths,

 

p M , is

1 1

i h M

i j

 

Packets may be lost due to channel error and queue overflow; in such cases, sending multiple packets on multiple paths will improve the reliability or reduce PER

Fig 8 is a specific example for the mathematical PER evaluation of single-path and multipath routing with a per hop channel error rates of 1% As we can see, the higher the number of paths the better the reliability, and the larger the number of hops, the lower the reliability or the higher the PER