DELAY EFFICIENT MAC PROTOCOL FOR DIFFUSION BASED

ROUTING IN WIRELESS SENSOR NETWORKS

Amir Yahyavi, Hamid Khalili and Nasser Yazdani

Electrical and Computer Engineering Department, University of Tehran, North Kargar St., Tehran, Iran

Keywords: MAC Protocol, Diffusion Routing, Delay Efficiency, Energy Efficiency, Energy Aware Routing, Load

Balancing, Wireless Sensor Networks.

Abstract: In this paper we present DESMAC a contention based Medium Access Control protocol for Diffusion based

routing in Wireless Sensor Networks. One of the main challenges in WSNs is to balance delay efficiency

and energy consumption. Surveillance and monitoring application as well as many other need low latency

data delivery; but, since sensor nodes have a small source of energy usually Active/Sleep cycles are used to

reduce the energy consumption which causes higher delay. We use routing information to adaptively change

the duty cycle for different loads. In our Cross Layer Design the Routing Layer can manipulate the duty

cycle of underlying MAC protocol. The diffusion control messages are used to adapt the duty cycle to

variation in the load. Also extensive use of some nodes can damage the connectivity of network. Therefore

we provide a mechanism to balance the load between several possible paths. We discuss DESMAC design

and compare our simulation results to S-MAC and IEEE 802.11 standard. DESMAC achieves significant

latency reduction (up to 50 times better delay than S-MAC) while ensuring energy efficiency and load

balanced delivery.

1 INTRODUCTION

The goal of a wireless sensor network is reliable data

reporting and minimum energy consumption. Sensor

nodes have a small supply of energy which makes

power management a great criterion in protocol

design for this kind of networks. Motes can last

approximately around 100-120 hours on a pair of

AA batteries in active mode (Kumar et al., 2006)

and battery capabilities only double every 35 years.

Several methods have been proposed to put

sensor nodes to inactive mode when the application

of the sensor network is able to tolerate the effects

(Ye, Heidemann and Estrin, 2004) (van Dam and

Langendoen, 2003) Putting nodes to inactive mode

reduces the idle listening time which is one of the

major sources of energy wastage. On the other hand

it causes higher delay, lower coverage & lower

connectivity in exchange for energy efficiency.

Since many applications have delay or coverage

criterion, use of many sleep scheduling schemes can

be limited (Akyildiz et al., 2002).

Medium Access Control protocols in wireless

sensor networks need to be energy efficient since it

is usually impossible to replace or recharge the

batteries. Moreover many applications have latency

and data rate requirements. Balancing the trade-off

between energy efficiency and delay efficiency has

been subject of great deal of research. Adaptive

adjustment of MAC layer’s duty cycle based on

current load and application needs is one of the

possible solutions. MAC layer’s information about

the network is usually not sufficient and vague for

example MAC layer usually does not have the

information necessary to determine if degradation of

data transmission is due to a neighbor’s failure or a

temporary distortion, etc. One of the ways to meet

both the energy and delay criterion of applications is

by using the help of Routing and Application layer

information to optimize the duty cycle of the MAC

layer based on the current load and situation.

Diffusion based routing algorithms are very

popular in wireless sensor networks

(Intanagonwiwat et al., 2003), (Ganesan, et al.,

2001). Routes are created on demand when an event

occurs and are not maintained by all sensors all the

time. All nodes in diffusion based routing are

application-aware. Directed Diffusion’s messages

can be used for adjustment of MAC layer’s duty

cycle to provide delay efficient data transfer.

Yahyavi A., Khalili H. and Yazdani N. (2008).

DELAY EFFICIENT MAC PROTOCOL FOR DIFFUSION BASED ROUTING IN WIRELESS SENSOR NETWORKS.

In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 88-91

DOI: 10.5220/0002024800880091

 SciTePress

2 DESMAC

We introduce a novel cross layer design to reduce

latency caused by periodic sleeping of the nodes.

Routing and application are used to help the MAC

protocol better adjust itself. We use S-MAC as the

basic framework for our design and adaptively

change its duty cycle. We assume that the reader is

familiar with S-MAC and Directed Diffusion (for

further reading please refer to (Ye, Heidemann and

Estrin, 2004) and (Intanagonwiwat et al., 2003)).

In this approach we use routing layer control

messages to adaptively reduce the sleep period of

the MAC layer of the nodes that take part in the

routing process. We decrease the sleep period of the

nodes that are reinforced by Directed Diffusion for

routing purposes. Nodes that are not reinforced

preserve their current sleep schedule.

In Directed Diffusion sink periodically sends

interest messages to all nodes in the network. When

a node detects an event which matches the diffused

interests it becomes a data source and starts sending

exploratory data messages. These messages are

forwarded towards the sink. When the sink receives

a positive reinforcement it issues a positive

reinforcement.

Positive reinforcements are similar to interest

messages but have lower interval. Each node

compares this interest message with the fields in its

own cache and if a lower interval is detected, it

updates its gradient toward that node to the new

value. In our approach positive reinforcements also

trigger a sleep period reduction in the MAC layer.

Negative Reinforcements have the exact opposite

effect on the sleep period. Negative reinforcements

in Directed Diffusion are used to reduce the number

of reinforced paths and path repairs. Several

mechanisms for negative reinforcement are

introduced in directed diffusion (timeouts, gradient

reductions, etc). Any mechanism for negative

reinforcement used by Directed Diffusion also

triggers an increase in the sleep period of the MAC

layer.

The goal is to benefit from energy saving

features of S-MAC and have much lower latency in

comparison to it. Nodes that are not reinforced by

the Directed Diffusion have the same duty cycle as

S-MAC, But nodes that are reinforced increase their

active period exponentially therefore upon path

establishment nodes on the path are almost always

active which results in very low latency that is

comparable to IEEE802.11.

Nodes that are not part of routing preserve their

original duty cycle and have similar energy

consumption to S-MAC. But routing nodes have

high duty cycle which provide very low latency and

can meet the application criteria.

In case a path failure or degradation occurs, the

path is negatively reinforced by Directed Diffusion

which will reduce the duty cycle of the nodes

previously involved in the routing. Therefore nodes

that are no longer involved in routing have low duty

cycle and energy saving is maximized.

S-MAC works on the basis that neighbor nodes

wake up at the same time therefore they can hear

each other’s broadcast messages (SYNC, RTS/CTS).

If neighbor nodes don’t have synchronous schedules,

communication between them becomes impossible.

Dynamic reduction of sleep period can disrupt

the synchronization done in the SYNC period of S-

MAC.

To address this problem we reduce the sleep

period in a manner that the SYNC and listen period

of the new schedule is still synchronized with that of

neighbor nodes. In order to achieve synchronized

wakeups we increase the duty cycle exponentially.

The sleep period in S-MAC is much longer than the

listen period therefore it is possible to reduce it so

that the frame size is divided in half. Each frame

turns into two frames with SYNC, RTS/CTS, and

sleep period which means that each positive

reinforcement message doubles the duty cycle of the

node until it achieves maximum possible duty cycle

This is shown in Figure 1.

Figure 1: Change in Duty Cycle as a node receives a

reinforcement message. Beginning of Listen periods is still

synchronized.

Increasing the duty cycle in this fashion does not

disrupt the synchronized wakeup of the neighbor

nodes. Nodes that are on the same path have the

same duty cycle. These nodes wakeup more often

and have more time for transmitting data therefore

provide lower latency and higher throughput. These

nodes are still able to communicate with the nodes

that have different duty cycles. Since the duty cycle

is increased exponentially the neighbor nodes still

have the synchronized wakeup. Nodes active in the

routing process have very small sleep periods and

show similar behavior to 802.11. But nodes not

involved behave similar to S-MAC.

Listen

Sleep

Listen

Time

Listen

Sleep

Sleep Listen Listen

Sleep Sleep

Before

After

Time

DELAY EFFICIENT MAC PROTOCOL FOR DIFFUSION BASED ROUTING IN WIRELESS SENSOR NETWORKS

3 LOAD BALANCING

Nodes that are involved in routing for a long period

will fail faster since they have higher energy

consumption as a result of higher duty cycle. Failure

of these nodes can cause the network to be

partitioned and other unwanted side effects such as

reduced coverage and connectivity.

In dense sensor networks, multiple paths may

exist between a source and destination; therefore, it

is desirable to use all these paths to efficiently use

the energy in a distributed manner. (Ganesan, et al.,

2001) is an example of a diffusion based multipath

routing algorithm. Directed Diffusion’s nature

makes it a good candidate for multipath routing

since initially reinforces several paths and then tries

to reduce the number of paths by using negative

reinforcements. In our cross layer design we use this

multipath potential of Directed Diffusion to adjust

MAC layer’s duty cycle to balance the network load

based on remaining routing nodes’ energy. We

define a critical remaining energy limit for the

nodes. When a node hits this critical limits it will

dramatically reduce its duty cycle which will

degrade the data delivery rate to the sink node. This

degradation is detected by Directed Diffusion and

triggers the local path repair mechanism. In local

repair mechanism a chain of local interactions result

in another path establishment that does not contain

the node that has reached the critical energy limit. In

a dense network where there are several neighbor

nodes that can replace a node with low energy level

this mechanism can be used to efficiently distribute

energy consumption on neighboring nodes.

In order to provide more control over different

energy levels we define several duty cycle reduction

levels. When a node reaches the predetermined

critical energy level it reduces its duty cycle to half

the original duty cycle in the deployment time. To

maintain the synchronization between sensor nodes

duty cycle reduction is exponential (similar to path

reinforcements). Every time a node reaches a new

energy level it will trigger a duty cycle reduction in

the same manner. For example if three energy levels

are defined the duty will be reduced up to one eight

of the original duty cycle. The critical energy limit

and different energy levels can be defined based on

the network, traffic, and application characteristics.

These levels can also be dynamically changed based

on current network situation and application needs.

4 SIMULATIONS

We compare our MAC protocol to 802.11 and S-

MAC. We used NS-2 for our simulations. Latency

and energy consumption of a Directed Diffusion

application on three MAC protocols: 802.11, S-

MAC, Delay Efficient S-MAC (DESMAC) will be

compared. The initial energy of the nodes is 3000

joules. To compare the energy consumption of each

protocol we use the power consumption model of

Cabletron 802.11 network interface card in

Transmit, Receive, Idle, and Sleeping modes (Chen,

et al., 2002). Nodes’ deployment is grid and source

and sink are at the ends of grid’s diagonal. Some

simulation results are not included due to space

limitations.

Figure 2 compares the latency between S-MAC

and DESMAC and IEEE802.11. As shown average

delay in DESMAC is much lower than S-MAC.

Since S-MAC is not able to transfer data message

with the required rate, queuing delay makes the

behavior of S-MAC unpredictable. It can also trigger

unwanted path changes in Directed Diffusion since it

may falsely detect degradation in the current path

because of high jitter. DESMAC on the other hand

shows a much more stable behavior. Small path

changes because of load balancing mechanism

create a small variation in DESMAC’s delay. As

expected delay of IEEE802.11 is lower than

DESMAC (always near zero).

Figure 2: Comparison of Packet delivery delay in different

MAC protocols for

grid deployment.

Figure 3 compares the energy consumption of S-

MAC, DESMAC, and IEEE802.11. IEEE802.11 has

the highest energy consumption since nodes are

always active and nodes fail much sooner than other

protocols (mid-way during simulation). DESMAC

WINSYS 2008 - International Conference on Wireless Information Networks and Systems

has higher energy consumption than S-MAC

because nodes involved in routing have higher

energy consumption than nodes with the normal

duty cycle.

Figure 3: Comparison of energy consumption in different

MAC protocols.

Figure 4 compares the average remaining energy

of the network.

Figure 4: Average network remaining energy for different

network sizes.

IEEE 802.11 as expected has the highest energy

consumption. Nodes are always active and the

number of nodes involved in routing doesn’t affect

the average total remaining energy of the network.

S-MAC has the lowest energy consumption, and

similar to IEEE 802.11 average remaining energy

doesn’t depend on the number of involved nodes in

routing. DESMAC’s total remaining energy is

somewhat close to S-MAC but since nodes involved

in routing have higher energy consumption total

remaining energy decreases as the number of these

nodes grow. The network size has another effect on

the average remaining energy. As number of paths

increases multiple path changes becomes possible

and different nodes during simulation may become

involved in routing therefore number of nodes that

have very high duty cycle decreases. This results in

lower energy consumption in network. But as

network size grows further number of nodes

involved in routing becomes a high percentage of

the deployed nodes therefore total remaining energy

of the network decreases.

5 CONCLUSIONS

In this paper we presented DESMAC, a delay

efficient MAC protocol for diffusion based routing

in wireless sensor networks. It supports power

saving features and adapts to data transmission load

in different situations. DESMAC does not pose any

messaging overhead for its adaptive duty cycling

(Since the control messages of Directed Diffusion

are used) and load balancing. It has much lower

delay in data transmission in comparison to S-MAC

and has much better energy consumption in

comparison to IEEE 802.11. In order to avoid failure

of nodes involved in routing due to higher energy

consumption in these nodes DESMAC changes the

path when these nodes hit a critical energy limit.

This results in higher network longevity and

preserving of network connectivity.

REFERENCES

Akyildiz, I.F. et al., 2002. Wireless sensor networks: a

survey. Computer Networks, 38(4), p.393-422.

Chen, B. et al., 2002. Span: An Energy-Efficient

Coordination Algorithm for Topology Maintenance in

Ad Hoc Wireless Networks. Wireless Networks, 8(5),

p.481-494.

Ganesan, D. et al., 2001. Highly-resilient, energy-efficient

multipath routing in wireless sensor networks. ACM

SIGMOBILE Mobile Computing and Communications

Review, 5(4), p.11-25.

Intanagonwiwat, C. et al., 2003. Directed diffusion for

wireless sensor networking. IEEE/ACM Trans.

Networks, 11(1), p.2-16.

Kumar, S., Lai, T.H. & Balogh, J., 2006. On k- coverage

in a mostly sleeping sensor network. Wireless

Networks.

Van Dam, T. & Langendoen, K., 2003. An adaptive

energy-efficient MAC protocol for wireless sensor

networks. Proceedings of the first international

conference on Embedded networked sensor systems,

p.171-180.

Ye, W., Heidemann, J. & Estrin, D., 2004. Medium access

control with coordinated adaptive sleeping for wireless

sensor networks. IEEE/ACM Trans. Netw., 12(3),

p.493-506.

DELAY EFFICIENT MAC PROTOCOL FOR DIFFUSION BASED ROUTING IN WIRELESS SENSOR NETWORKS