You are on page 1of 4

Energy-aware Routing for e-Health Wireless Sensor Networks

Telemedicine, e-health and telecommunications

Abstract Wireless monitoring devices are an ideal option for


e-health solutions because of their usage convenience. One of
the most important challenges in the design of a wireless sensor
network is to extend the lifetime of the monitoring devices.
The solution presented in this paper is focused on providing a
dynamic routing strategy for wireless sensor networks used in ehealth systems, with the main objective of increasing network
lifetime. We use a common network model and we propose
a novel energy aware routing algorithm that discovers and
selects the best route between a source and the sink node
in a distributed fashion. We selected the Weighted Product
Model (WPM) as the mechanism used to estimate the cost of
the routes. The energy consumption for the proposed algorithm
is evaluated with a target platform (e.g. mote - sensor node).
Keywords: e-health, wireless sensor networks, routing algorithm, energy-awareness

I. INTRODUCTION
It is already known that world population is ageing. This
happens because of scientific and technological de development which increase the life expectancy at birth [1] by
almost 20 years (from 46.5 years in 1950-1995 to 66.0 years
in 2000-2005) and second because fertility dropped almost to
half, from 5.0 to 2.7 children per women [2]. By 2050 the
aged population will increase by 70 percent in developed
regions and by 400 percent in less developed regions [2].
One of the healthcare strategies for ageing society is the
use of technology to provide care to ageing population [3] in
their own environment at home or at elderly care institutions.
This is possible through the use of telehealth solutions such
as remote patient monitoring or telemonitoring.
Wireless monitoring devices represent an attractive option
for telehealth solutions because they do not restrict the
movement of the patient (i.e. lightweight and small form
factor). There are several approaches [4][5][6][7] concerning
such monitoring systems that make use of commercially
available hardware and communication stacks. The use of
low data rate and low power short range radios introduces
the need of data routing among such networks. The spatial
placement of the nodes and rigid routing strategies are the
two main factors that determine uneven energy depletion.
This problem tends to scale with the network size as some
nodes become rapidly unavailable due to the high amount of
traffic that is routed trough them.
Our work is focused on providing a dynamic routing strategy for wireless sensor networks used in e-health systems,
with the objectives of balancing the routing load throughout
the network and increasing network lifetime (i.e. average
node operating time). Also, we chose to implement this
solution on 2.4 GHz IEEE 802.15.4 compliant devices due to
their wide availability from different manufactures and due
to their low-power capabilities.

II. RELATED WORK


The literature includes several algorithms for WSNs.
Although they were not designed specifically for medical
WSNs, their goals are related to reduced energy consumption
or increased network lifetime.
In [8] the authors address the problem of uneven energy
depletion by introducing a distributed energy balanced routing (DEBR) algorithm. Based on a new metric, transmission
energy cost, the nodes decide if data is to be transmitted
to the base station directly or through peers. Experimental
results show that DEBR is robust to diverse event generation
patterns and prolongs network lifetime compared to direct
communication. Nevertheless, the algorithm assumes that
the wireless radios have several transmission power levels,
including one that can reach the base station, which is not
always true in wireless sensor networks.
In [9] and [10] the problem of energy-efficient routing
is treated as a network formation one. A IEEE 802.15.4
compliant PAN coordinator ELection procedure (PANEL)
is used to balance all the routes in the network. This
approach minimizes the average number of hops between
the nodes in the network and the PAN coordinator leading
to reduced data transfer delay and energy savings. However,
PANEL requires all the nodes in the network to support
coordinator functionality which yields a higher cost. Also,
the node residual energy is not taken into account in the
reconfiguration phase which may lead to broken links early
in the network life.
The power-aware routing approach in [11] uses a novel
link metric which attains the best trade-off between power
consumption of the node and its hop-distance to the sink.
Based on the Djikstras algorithm, it finds the optimal routing
path between all the sensor nodes in the network and the
sink. Simulation results show little improvement in power
consumption compared to hop-count routing and residual
energy routing.
III. MODEL AND ASSUMPTIONS
In this section, we describe the wireless network communications model, concerning topology, packet routing and
energy model.
A. Network Model
This work assumes a wireless mesh network composed
of N nodes, randomly deployed in a two-dimension field.
Similar to papers [11]-[14] we consider that all nodes are
homogenous (i.e they can perform the same functions regarding our algorithm) and that all data is forwarded to a
single sink node. Also, the links are symmetric meaning that

TABLE I
S AMPLE ROUTING TABLE

data transmission either way consumes the same amount of


energy.
B. Routing Model
Route discovery is based on a modified version of AODV
[15].
The AODV routing protocol uses an on-demand approach
for finding routes (i.e. the network is silent until a connection
is needed). When trying to find a route, the source node
broadcasts a RouteRequest throughout the network. When
the request reaches a node that has a route to the destination,
it broadcasts a RouteReply. The RouteReply propagates to
the source node and adds the routing information to the
nodes in-between. All AODV packets have a timestamp (i.e.
destination sequence number) that is used to refresh the
entries in the AODV routing tables that are stored locally. In a
single enquery session the nodes may obtain the latest routes
to different destinations, but only one route per destination.
Usually, this route is the shortest path (i.e. minimum number
of hops) between the source and destination.
Our implementation allows the nodes to store multiple
route information for one destination which is needed for
our methodology. The fields of a table entry are presented
in Table I. Dest represent the destination node and Next
represent the next node in that route (it can also be the
destination). No hops holds the number of nodes that are
in between the destination and the current node. If there are
more routes to the destination that have different lengths,
but have the same next node, the table will hold an entry
for each. Min energy field holds the value of the minimum
residual energy along the route. Upon reception of routing
information the Timestamp field is used to determine if the
current table entry needs to be updated. Current entries in the
local routing tables are discarded if a node along the route
leaves the network and are regularly updated with the minimum residual energy information. The RouteRequest and
RouteReply packets are modified to carry this information.
In order to keep an up-to-date table, the residual energy
information is also propagated by the data and acknowledge
packets.
Table I also presents some sample entries for routes to
nodes 5 and 10. First two entries show routes to node 5 that
both pass trough node 7 but have different lengths. Last two
entries show routes to node 10 that have a different next node
but have the same length. If there are two or more routes that
have the same next node and the same length, the Min energy
field will hold the maximum value of the minimum energy
along each route.

Dest
5
5
10
10

N ext
7
7
3
4

N oHops
4
5
3
3

M inEnergy
4.57 J
2.21 J
5.02 J
2.54 J

Require: RT able
. Local routing table
Require: p
. Received routing packet
Require: id
. Node ID
Require: REnergy
. Residual energy of current node
1: procedure R ECV ROUTE R EQUEST
2:
Update/Add rt RT able where rt.N ext = p.Src
and rt.Dest = p.Origin
3:
if p.Dest = id then
4:
Generate RReply
5:
return
6:
end if
7:
rts RT able.getRoutes(p.Dest)
8:
if rts.Count() > 0 then
9:
if more than one route with same N oHops then
10:
mE min(max(M inEnergy), REnergy)
11:
Generate RReply with mE
12:
else
13:
Generate RReply
14:
end if
15:
else
16:
mE min(p.M inEnergy, REnergy)
17:
Generate RouteRequest with mE
18:
end if
19: end procedure
Fig. 1.

Pseudocode - Routing packet reception

A. Route Discovery and Update


The route discovery process is started when a node
does not know how to reach a peer and broadcasts a
RouteRequest. These packets propagate throughout the network until they reach the destination node or a node that has
a route to the destination. Next, a RouteReply is propagated
back to the node that requested the route discovery. Fig. 1
presents the actions taken upon receiving of a RouteRequest
packet.
First, an entry is added or updated in the routing table
for the route back to the node that generated the first
request. If the node is the intended destination, it sends a

IV. ENERGY-AWARE ROUTING


The routing algorithm is intended to discover and select
the best route between a source and the sink node in a
distributed fashion. If there are multiple paths towards the
sink, a node uses a cost function to select the next hop with
the objective of prolonging the network lifetime. This process
in based on two activities: Routediscoveryandupdate and
Costcalculation.

T imestamp
54
45
60
61

Fig. 2.

Routing example

RouteReply and idles afterwards. Next, if no route is found,


the node forwards the RouteRequest with updated energy
information.
If the destination is not reached, the node searches its
routing table and for every route found with a distinct
N oHops a RouteReply is sent. If there is more than one
route with the same number of hops, the node determines the
maximum value of the M inEnergy field. These routes are
transparent to the nodes closer to the source (i.e. they dont
know how many they are) and only best route information
has to be sent back (i.e. if two routes have the same length,
only the M inEnergy criteria is considered in the decision
process).
An example is presented in Fig. 2. The source node (S)
has three routes to the destination (D), but is aware only of
two. In the route discovery process, node R reports back to S
only one route, because it has two possibilities with the same
N oHops. If node S decides to forward the data packets to R,
R will then choose the best of the two available routes. But
in order to decide whether to forward the data to R or to send
it along route3, S needs to know information about the best
route that passes through R. That is why, the M inEnergy
information passes by R to S is the one for route2.
A similar process takes place when receiving a data packet.
The node updates its routing table with the return route
information from the reply packet (i.e. the communication
process includes acknowledgement of data reception). If the
packet has to be forwarded, the M inEnergy field is updated.
If there are more return routes with the same number of hops,
the node searches the one with the highest minimum residual
energy as described in the previous paragraph.
B. Cost Function
We selected the Weighted Product Model (WPM) as the
mechanism used to estimate the cost of the routes.
Given a set of m alternatives to be selected, a set of
n decision criteria and wj , j = 1, n, the relative weight
of importance for criterion Cj , the performance value of
alternative Ax relative to alternative Ay is calculated as in
(2).

P

Ax
Ay


=

n
Y

ajx

j=1

ajy

!wj
, x = 1, m, y = 1, m

(1)

where aji is the performance value


P of alternative Ai
evaluated in terms of criterion Cj and
wj = 1.
If the ratio P (Ax /Ay ) 1, then the alternative Ax is said
to be more desirable than alternative Ay , when the considered
decision problem is a maximization one. In the case of a
minimization problem, the ratio should be less than one for
Ax to be more desirable than Ay .
The WPM has two main advantages when considering our
problem: it has a low implementation complexity, expressed
as processing overhead, and it is a dimensionless analysis
method, meaning it eliminates any units of measure from
the performance values of the alternatives.

Fig. 3.

Evaluation scenario network setup

In particular, the elements considered for the WPM are


the following: the alternatives are represented by the routes;
the evaluation criteria are represented by the N oHops,
M inEnergy characteristics of the routes and by the length
of the data packet. The fine tuning of this mechanism is done
by adjusting the weights of the three criteria.
C. How it works
If a source node does not have a route to a destination it
start a discovery similar to AODV but using the described
modifications. When more than one route is available for
a given destination, the node calculates the relative ratios
between them, two-by-two, using WPM. The resulting set is
then ordered based on the relative cost and the route with
the lowest cost is chosen.
V. EVALUATION
We chose to evaluate our energy-aware routing algorithm
on the network topology presented in Fig 3.
The simulated network is composed of 11 nodes plus the
sink with a transmission range of 20 meters. The nodes are
arranged in such way, in a bi-dimensional area, that only the
connections in Fig. 3 exist between them.
First, we ran the route discovery process. If we consider
only the route from node 11 to the sink node, we see that
there are eight routes (Table II). As expected, the routing
table of node 11 was built as in Table III. The information
available to node 11 is limited to the information about its
neighbours. The fact that both r1, r2 pass trough node 4 and
r3, r4 pass trough node 10 is transparent to node 11. An
interesting observation can be made about these routes: as
the table entries do not have the M inEnergy field equal
and the smaller value is for the route through node 10, we
can conclude that this value corresponds to node 10. More,
we see that there are two routes trough node 9 that have
different lengths. The same value for the M inEnergy field
suggests that it corresponds to a common node (nodes 9, 7,
5, 3).

Next, we evaluated the energy consumption for the proposed algorithm. The target platform is based on a low-power
EFM32 ARM Cortex-M3 MCU and an IEEE 802.15.4 2.4
GHz MRF24J40 radio. The platform is equipped with temperature and accelerometer sensors and can support extension
boards for SpO2, EKG and other physiological signals.
The test application states and parameters are shown in
Table IV. The Battery Estimator and Energy Profiler tools
from Energy Micro [16] were used to estimate the system
operating time. For a rated battery capacity of 3000 mAh the
estimated operating time is 264 days with an average current
consumption of roughly 500 uA.
TABLE II
N ODE 11 S INK ROUTES
Route
r1
r2
r3
r4
r5
r6
r7
r8

Nodes
11-4-2-S
11-4-1-S
11-10-4-2-S
11-10-4-1-S
11-8-6-5-3-S
11-8-7-5-3-S
11-9-7-5-3-S
11-9-7-8-6-5-3-S

TABLE III
N ODE 11 ROUTING TABLE
Dest
S
S
S
S
S

N ext
4
10
8
9
9

N oHops
3
4
5
6
5

M inEnergy
1.47 J
1.18 J
2.01 J
1.22 J
1.22 J

T imestamp
11
12
5
19
17

TABLE IV
T EST APPLICATION STATES
State

Description

Current

TX

radio module in TX state, MCU


acquiring data, selected peripherals
active (i.e. sensor serial interfaces,
ADC, DAC)
radio module in RX state, MCU processing received data, selected peripherals active (i.e. timers, storage)
radio module in Sleep state, MCU
in low-power mode, interrupt system
active

29.47 mA

Duty
cycle
0.1%

21.80 mA

2%

590 nA

97.9%

RX

Idle

VI. CONCLUSIONS
In this paper we have proposed a novel routing algorithm
for wireless sensor networks. The algorithm chooses the
communication routes dynamically in order to balance the
energy consumption throughout the network. This makes
it suitable for deployment in patient monitoring systems
because it can extend the lifetime of the monitoring devices.

The evaluation of the algorithm concluded that it achieves


the routing objectives and that it is suited for battery powered
devices.
We intend to further investigate the energy consumption
and network capacity of the algorithm by deploying a network of nodes and monitoring their parameters.
R EFERENCES
[1] W.C. Mann, The aging population and its needs, IEEE Pervasive
Computing, vol. 3, pp. 12-14, 2004.
[2] UN, Department of Economic and Social Affairs, Population
Division, World Population Ageing: 1950-2050 [Online]. Available:
http://www.un.org/esa/population/publications/
worldageing19502050/.
[3] The Economist, Economist Intelligence Unit, Healthcare
strategies for an ageing society [Online]. Available:
http://graphics.eiu.com/upload/eb/Philips_
Healthcare_ageing_3011WEB.pdf.
[4] C. Rotariu, H. Costin, D. Arotaritei and B. Dionisie, A Low Power
Wireless Personal Area Network for Telemedicine, in Proceedings of
the European Congress on Biomedical Engineering, 2008, pp.982-985.
[5] L. Boquete et al., A portable wireless biometric multi-channel system, Measurement, vol. 45, issue 6, pp. 1587-1598, 2012.
[6] M.C. Batistatos, G.V. Tsoulos and G.E. Athanasiadou, Mobile
telemedicine for moving vehicle scenarios: Wireless technology options and challenges, Journal of Network and Computer Applications,
vol.35, issue 3, pp. 1140-1150, 2012.
[7] C. Rotariu, A. Pasarica, H. Costin, F. Adochiei and R. Ciobotariu,
Telemedicine system for remote blood pressure and heart rate monitoring, in Proceedings of the E-Health and Bioengineering Conference (EHB), 2011, pp. 1-4.
[8] C.-S. Ok, S. Lee, P. Mitra and S. Kumara, Distributed energy balanced routing for wireless sensor networks, Computers & Industrial
Engineering, vol. 57, pp. 125-135, 2009.
[9] F. Cuomo, A. Abbagnale and E. Cipollone, Cross-layer network
formation for energy-efficient IEEE 802.15.4/ZigBee Wireless Sensor
Networks, Ad Hoc Networks, vol. 11, pp. 672-686, 2013.
[10] F. Cuomo, E. Cipollone and A. Abbagnale, Performance analysis of
IEEE 802.15.4 wireless sensor networks: An insight into the topology
formation process, Computer Networks, vol. 53, pp. 3057-3075,
2009.
[11] E. Sanchez, L. Murillo, B. Montrucchio and M. Rebaudengo, An
Adaptive Power-Aware Multi-hop Routing Algorithm for Wireless
Sensor Networks, in Proceedings of the Eighth International Conference on Information Technology: New Generations, Las Vegas, 2011,
pp. 112-116.
[12] S. Tyagi and N. Kumar, A systematic review on clustering and routing
techniques based upon LEACH protocol for wireless sensor networks,
Journal of Network and Computer Applications, vol. 36, pp. 623-645,
2013.
[13] B. Lee, S. Bae, and D. Han, Design of Network Management
Platform and Security Framework for WSN, in Proceedings of the
2008 IEEE International Conference on Signal Image Technology and
Internet Based Systems, Bali, 2008, pp. 640-645.
[14] H. Gou, G. Li, and Y. Yoo, A Partition-Based Centralized LEACH
Algorithm for Wireless Sensor Network Using Solar Energy, in
Proceedings of the International Conference on Convergence and
Hybrid Information Technology, Daejeon, 2009, pp. 60-66.
[15] C. Perkins, E. Belding-Royer and S. Das, Ad hoc On-Demand
Distance Vector (AODV) Routing, IETF RFC 3561, 2003.
[16] T. Monte, Energy Debugging Tools for Embedded Applications,
white paper, Energy Micro AS, 2010.

You might also like