Wireless Sensor Network, 2011, 3, 106-113
doi:10.4236/wsn.2011.33011 Published Online March 2011 (http://www.SciRP.org/journal/wsn)
Energy-Efficient and Reliable Transport Pr otocols for
Wireless Sensor Networks: State-of-Art
Ahmed Ayadi
Institut Telecom/Telecom Bretagne, Cesson-S évigné, France
Received February 14, 2011; revi sed February 28, 2011; accepted March 8, 2011
Abstract
New wireless sensor network applications (e.g., military surveillance) require higher reliability than a simple
best effort service could provide. Classical reliable transport protocols like Transmission Control Protocol
(TCP) are not well suited for wireless sensor networks due to both the characteristics of the network nodes
(low computing power, strong energy constraints) and those of the main applications running on those nodes
(low data rates). Recent researches present new transport protocols for wireless sensor networks providing
various type of reliability and using new mechanisms for loss detection and recovery, and congestion control.
This paper presents a survey on reliable transport protocol for WSNs.
Keywords: Wireless Sensor Networks, Transport Protocol, Reliability, Congestion Control, Energy
Efficiency
1. Introduction
Wireless sensor networks (WSNs) [1] have important
applications such as remote environment monitoring
(temperature and humidity) and target tracking (for mili-
tary purpose).
In monitoring area, a WSN is deployed over a region
where some phenomena (e.g., floods, fires) are to be
monitored. This has been enabled by the availability of
sensors that are smaller, cheaper and intelligent. Sensors
are equipped with low-power wireless interfaces, which
are using it to communicate with each other to form a
network.
Firstly, WSN applications are not interested in relia-
bility because WSNs have been considered as fault- to-
lerant networks where sensor nodes collect environment
information and send it to the base station (sink). Inter-
mediate nodes offer their best effort services to relay
generated packets to the base station. However, new
WSN applications like military applications (e.g., battle-
field surveillance) require more and more reliability.
Moreover, in [2], the authors present a need of re-tasking
/reprogramming sensor nodes, and thus a need to send a
binary file or a script file to sensor nodes.
A transport protocol for WSNs should be reliable (or
provides different levels of reliability for each kind of
applications) and energy efficient (reduces the amount of
exchanged messages to reduce total consumed energy
and thereafter increases the network lifetime). The relia-
bility requires two essential mechanisms: congestion
control (detection and avoidance), and loss detection and
recovery.
The congestion control mechanism is an essential
component for a reliable transport protocol because the
congestion leads to packet losses. Losses in WSNs are
not only due to congestion but also to bad channel errors,
collisions and interference. To distinguish between the
two types of losses, an explicit congestion no tification is
proposed. These two mechanisms can be implemented in
a distributed form (in sensor nodes) or in a centralized
form (in the based station).
In a distributed form, sensor nodes use the sequence
numbers of packets to signal packet losses. A gap in the
sequence numbers signals a loss. When a loss happens,
an intermediate node may request the retransmission of
the data from its neighbor nod es. Senso r nodes can d etect
congestion based on the buffers overflows and then slow
down their sending rate.
However, in centralized protocol (e.g., RCRT [3]), the
base station is the only node responsible for the packet
loss detection and recovery. The base station detects also
congestion in the network using the arrival time of the
out-of-order packet s.
This paper presents a survey of reliable transport pro-

A. AYADI
107
tocols for WSNs and identifies some challenge’s re-
search. The next section presents the needs of an ener-
gy-efficient and reliable transport protocol for wireless
sensor networks and presents the recent proposed trans-
port protocols for wireless sensor networks in the litera-
ture. Section 3 evaluates and compares different proto-
cols in term of reliability, congestion control and energy
efficiency. The last section concludes the paper.
2. Transport Protocol in WSNs
Transport layer ensures the reliability and the quality of
data at the sender and the receiver. Transport protocols in
WSNs should support multiple applications (industrial
process monitoring and control, machine health moni-
toring, environment and habitat monitoring, health care
applications, home automation, and traffic control) and
provide variable reliability level, packet loss recovery
and congestion control mechanism. We can distinguish
between two types of data in WSNs:
1) Data sent by sensor network to the base station
(Sink),
2) Data sent by the base station to one or a subset of
sensor nodes for different purposes (control, manage-
ment, re-tasking, reprogramming).
The development of a transport protocol should be
generic and independent. It should provide various relia-
bility levels for different app lications. WSNs suffer from
a high loss rate. Packet loss may be due to bad radio
communication, congestion, packet collisions, full mem-
ory capacity, and node mobility or fail. Thus, transport
protocol should provide two functions: reliable data
transport and congestion control. A reliable application
requires that all segments sent by a source arrive to the
destination. Missing segments that may be lost in the
WSN should be recovered by reliable schemes. Conges-
tion happens when the data packets generated by sensor
nodes exceed the network capacity. When the networks
get congested, intermediate sensor nodes may drop
packets. This leads to retransmissions of the dropped
packets and thus a waste of energy that is a very impor-
tant factor in wireless sensor network. In the following,
we present a summary of recent transport protocol pro-
posed for WSNs.
2.1. Reliable Multi-Segment Transport
F. Stann et al. present in [4] the first transport layer with
hop-by-hop recovery scheme using caching mode as ad-
ditional control traffic for Direct Diffus ion [5]. The main
goal of RMST is to minimize the cost of end-to-end re-
transmissions. RMST protocol provides two transmission
modes: caching mode (with hop-by-hop recovery) and
non-caching mode (with end-to-end recovery). In
non-caching mode, only sources and sinks maintain a
cache, and only sinks set timers to detects loss.
In caching mode, RMST protocol assumes that each
sensor node has a cache memory where recently received
segments can be saved. RMST protocol reduces end-to-
end retransmissions by introducing hop-by-hop retrans-
missions from caches of neighbor nodes. In link layer,
lost packets are retransmitted using Automatic Repeat
reQuest (ARQ) [6].
The RSMT receivers are responsible for detecting
losses and for trigger the recovery of the missing seg-
ments through the generation of Negative Acknowledg-
ments (NACKs). The RSMT receivers are not only sinks,
but also intermediate nodes. To handle losses, an RSMT
intermediate node should store data traffics and con-
structs a map of received segments. When an out-of-
order segment is received, an RSMT receiver sends a
NACK requesting the retransmissions of the lost mes-
sages. Firstly, the one-hop neighbors process NACKs.
Then, if one of the neighbors finds the missing segments
in cache, it suppresses the NACK message and retrans-
mits the missing segments to the sink. Else, the NACK
message is relayed to the next node toward the source.
RSMT protocol provides 100% reliability even for ap-
plications that do not require total reliability. Moreover,
RMST does not include real time guarantees or conges-
tion control mechanisms. Require that each node has
enough memory to store all received segments is a very
strong condition difficult to be satisfied with memo-
ry-constraint wireless devices.
2.2. Pump Slowly, Fetch Quickly
Pump Slowly, Fetch Quickly [2] mechanism is proposed
for re-tasking/re-programming a group of sensors over-
the-air. PSFQ is based on slowly injecting packets into
the network “pump operation” and performing aggres-
sive hop-by-hop recovery in case of packet losses “fetch
operation”. Like RSMT, PSFQ provides a hop-by-hop
error recovery mechanism in which intermediate nodes
take the responsibility of loss detectio n and recovery. To
enable a hop-by-hop loss recovery and in-sequence data
delivery, a data cache is created and maintained at inter-
mediate nodes.
The PSFQ “pump operation” consists in a timely con-
trolled data forwarding. In intermediate nodes, when a
packet is received in an out-of-order sequence, it is
stored. However, instead of forwarding it, the immediate
node requests retransmission of the missing segment.
The PSFQ “fetch operation” is a proactive act of re-
questing a retransmission from neighboring nodes once
loss is detected at the receiving node. It corresponds to

A. AYADI
108
sending NACK for a retransmission request containing
the sequence number of the missing segment. If the up-
stream neighbors do not posses the missing segment,
they forward the NACK further, until it reaches a node
having the missing segments.
However, the use of PSFQ for the forward direction
can lead to a waste of energy. Besides this, PSFQ does
not address packet losses due to congestion. PSFQ give
good performance in a chain scenario where a sensor
node has only two neighbors. Moreover, in randomly
distributed network, PSFQ could flood network in fetch
phase by NACK messages.
2.3. Event-to-Sink Reliable Transport protocol
ESRT [7] is a transport protocol that seeks to achieve
reliable event direction with minimum energy expendi-
ture and congestion resolution. The main goal is to con-
figure the reporting frequency rate to achieve the desired
event detection accuracy with minimum energy expend-
iture. In fact, a sensor node in ESRT sends messages
with an announced reporting frequency to the base sta-
tion. An ESRT base station regulates the reporting rate of
sensors in response to a congestion detected in the net-
work. Congestion control mechanism is implemented in
the base station, which informs all sensor nodes using a
different technology about the new reporting frequency.
An ESRT node monitor s its local buffer level and sets
a congestion notification bit in the packets it forwards to
base station if the buffer overflows. If the base station
receives a packet with the congestion notification bit set,
it broadcasts a control signal informing all source nodes
to slow down their common reporting frequency. The
ESRT base station must broadcast this control signal at
high energy so that all sources can hear it or use another
technology.
Such a signal has several potential drawbacks, howev-
er, particularly in large sensor networks. Any on-going
event transmission would be disturbed by such high-
powered congestion signal to sour ces. In addition, ESRT
always regulates all sources regardless the congestion
region. ESRT does not retransmit lost packets.
2.4. Distributed TCP Caching
In [8], Dunk els et al. present Distributed TCP Caching, a
new scheme for TCP [9] in multi-hop wireless networks
that uses segment caching and local retransmission in
cooperation with link layer for TCP/IP-based wireless
sensor network. DTC is an extension of Snoop [10] idea
towards multi-hop sensor networks.
The authors assume that each intermediate node is
able to cache a single TCP data segment. DTC relies
mainly on timeouts to detect packet losses. Thus, each
node measures the round-trip time (RTT) to the receiver
and adapts a retransmission timeout RTT to 1.5 x RTT.
The authors propose to compute the RTT in the TCP
connection setup phase and to use RTT = 1.5 RTT as a
timeout value. The sensor nodes cache the TCP segment
that has the highest segment number seen with a certain
probability (p = 50%). An unacknowledged packet in
link layer should be locked and retransmitted after the
timeout. Locked d ata segments shou ld not be overwritten
by a TCP segment with higher sequence number. A
locked segment is removed from the cache only when a
TCP ACK that acknowledges the cashed segment is re-
ceived, or when the segment times out.
DTC uses also TCP SACK option [11] to both packet
loss detecting and signaling mechanism between DTC
nodes. The TCP SACK option is used by sensor node to
inform other nodes about segments locked in their cach-
es.
2.5. TCP Support for Sensor Networks
In [12], Braun et al. present TCP Support for Sensor
networks (TSS), which is a new layer between TCP and
network layer. TSS requires storing state information for
each TCP connection that contains sequence numbers,
acknowledgments numbers, and RTT.
TSS uses Implicit ACK (IACK) for loss detection: a
sensor node is assumed to listen to packet transmission
of their neighbor to detect whether the next node have
forwarded TCP segment. A node usi ng TSS always cach-
es a packet until it is sure that the successor node towards
the destination has received the segment. The retrans-
missions are mainly triggered by timeouts, which re-
quires careful setting of timeout values. Like DTC [8],
the retransmission timeout is set to 1.5 RTT. To avoid
congestion, a TSS node should stop forwarding its pack-
ets until it knows that all earlier packets have been re-
ceived and forwarded by its successor node.
The simulation results show that TSS gives more
throughput than TCP and less exchanged messages than
DTC [8]. However, hearing all neighbor node traffics is
not energy-efficient because listening power is important
as well as transmission power. In addition, a message
transmission fail of one sensor node leads to stop the
transmission of all its previous sensor nodes.
2.6. Asymmetric Reliable Transport
Asymmetric Reliable Transport [13] is an asymmetric
and reliable transport mechanism that does not address
the reliability of event notifications (sensor-to-sink) but
also the queries (sink-to-sensors), being thus a bidirec-

A. AYADI
109
tional transport protocol. ARP classifies the sensor as
essential (E) nodes and non-essential (N) nodes. The E
nodes are selected periodically to cover the entire sensi-
ble terrain, based on residual energy.
Using both asymmetric acknowledgment and negative
acknowledgment provides the end-to-end reliable com-
munications. To distinguish the last query message of
sequence, a Poll (P)/Final (F) bit is used.
2.7. Distributed Transport for Sensor Networks
Distributed Transport for Sensor Networks [14] is an
energy-efficient hop-by-hop reliable transport protocol
using both ACK and NACK message for delivery con-
firmation. A DTSN node analyzes the sequence numbers
of received packets and detects losses by finding gaps.
Every source node sends an Explicit Acknow ledgment
Request (EAR) every one Acknowledgment Window
(AK) to ask for an ACK or a NACK. The sink node rep-
lays by an ACK message if no gap is detected or by a
NACK message containing the sequence numbers of
missing segments.
DTSN protocol is a hop-by-hop recovery protocol; all
intermediate node caches received packet in their cache.
Upon reception of an ACK message, intermediate node
deletes acknowledged segment. Otherwise (i.e. reception
of NACK message) an intermediate node checks if its
cache contains one of the missing segment.
DTSN node retransmits missing segments and updates
the NACK message. DTSN offers two types of service:
total reliability service and differentiated reliability ser-
vice. The difference between the two types of service is
the probability of caching a segment in an intermediate
node. For example, in full reliability scenario, all seg-
ments are cached in intermediate nodes. DTSN algorithm
does not threat congest i o n detection and control.
2.8. Wisden
Wisden [15] provides a reliable data transport from the
sensor nodes to the base station. In Wisden, nodes self-
organize themselves into a routing tree rooted at the base
station. Wisden implements both en d-to-end and hop-b y-
hop NACK based reliability scheme. Nodes keep a small
cache of recently transmitted packets. Intermediate node
detects packet loss based on gap in the sequence numbers
in a received segment. Entries in the “missing packets”
list are piggybacked in the outgoing transmissions, and
children infer losses by overhearing this transmission.
Lost packets are often recovered hop by hop, however,
two factors necessity end-to-end recovery: the large list
of missing packet that can exceed the memory of the
sensor node and the topology changes. Limit of Wisden
is the constant value of sending rate, which should be
measured and configured based on the bandwidth and the
number of nodes.
2.9. Rate-Controlled Reliable Transport protocol
Rate-Controlled Reliab le Transport protocol [3] is a mul-
tipoint-to-point reliable transport protocol for wireless
sensor networks. RCRT uses an explicit end-to-end loss
recovery and places all congestion detection, recovery
and rate adaptation schemes in the base station (Sink).
RCRT sink has three distinct logical components:
a) End-to-end retransmission
The main goal of RCRT is to achieve 100% reliability.
The RCRT sink uses NACK-based end-to-end loss re-
covery to request the retransmissions of missing packets
from the source. Each source (sensor node) has a re-
transmission buffer where is saved the not ack nowledged
segment. The sink node keeps a list of lost segments then
sends a NACK feedback message to the source contain-
ing the sequence numbers of missing segments. Upon
receiving a NACK, the source node retransmits the re-
quested segments.
b) Congestion detection
To distinguish between congestion and transmission
losses, RCRT congestion detection mechanism is based
on the length of the losses. The sink nod e maintains a list
of the out-of-order messages and computes the Time to
recover loss. If this value exceeds 2RTT, congestion
is then signaled.
c) Rate adapt ation
RCRT uses AIMD to adapt the transmission rate of
each source. Whenever the RCRT sink determines the
network is congested, it applies the rate decrease and
computes the new rate for all flows:
 Increase:


1RtRtA

,
 Decrease:


1RtMt Rt ,
where A is a constant and M (t) is a function of loss rate,




2
i
i
pt
Mt pt and


i
ptis the loss rate value of the
source i at the instant t.
2.10. Interference-Aware Fair Rate Control
Interference-aware Fair Rate Control [16 ] is a distributed
rate allocation scheme that uses queue size to detect
congestion, shares congestion state through overhearing
messages and converges to fair and efficient rates for
each node. IFRC scheme consists of three components:
a) Measure of level congestion
An IFRC node uses an exponentially we ighted mov ing
average of instantaneous queue length as a measure of

A. AYADI
110
congestion

1
qqqqq
avgwavgw inst .
The average packet length is updated whenever a
packet is inserted into the queue. IFRC detects incipient
congestion by using multiple thresholds

Uk where
 
1
12k
I
Uk Uk  where k is a small integ e r
and
I
is a constant increment of queue length.
b) Congestion Sharing
In IFRC node i includes the following information
in the header of each outgoing transmission packet: its
rate i r, current average queue length, a bit indicating
whether any child of i is congested, small state i r
among all its congested children, and l’s average queue
length. IFRC introduces two simple constraints:
 Rule 1: i r can never exceed
j
r, the rate of i’s
parent j.
 Rule 2: Whenever a congested neighbor j of i
crosses a threshold


Uk, i sets its rate to the
lower of i r and l r.
The same rule is applied for the most congested child
l of any neighbor of i, i sets its rate to the lower of
i
r and
j
r where l is the most congested ch ild of i’s
neighbor.
c) Rate Adaptation
All nodes start from a fixed rate init r. IFRC imple-
ments multiplicate rate increase initially. After slow start
phase, an IRFC node increases it rate i
r every 1i r. If
node i is congested, then when threshold


Uk is
crosses, the node halves it current rate. The base station,
even if it does not source messages, it maintains its
“rate” h
r and adapts using the same mechanism de-
scribed bellow. Because the base station does not send
messages, it broadcasts a control message after the re-
ception of 5m messages to share its rate h
r. IFRC
presents a shared congestion control mechanism but not
reliable. Moreover, IFRC adds a lot of overhead in the
header of transport protocol stack and a significant
amount of control messages.
2.11. Energy-efficient and Reliable Transport
Protocol
ERTP [17] is an energy-efficient and reliable transport
protocol for low data streaming in WSNs. ERTP is a
hop-by-hop recovery algorithm using Implicit Acknowl-
edgment. ERTP requires that each node i after sending
a packet to the next node to the sink overheads the next
forwarding. The forward of a packet by node 1i

is
considered as an implicit acknowledgment to node i.
The authors present a hop-by-hop reliability control,
which adjusts the maximum number of retransmission of
a packet in each node based on the link loss rate. They
present also an algorithm for computing the extents time
in witch node i is expected to “overhead” the forward-
ing packet of node 1i

.
In the result section, authors show that the use of
ERTP algorithm for computing the RTO time is better
that Jacobson algorithm. They show also that using
ERTP gives more delivery ratio than using simple expli-
cit acknowledgment. However, hearing all neighbor node
traffics is not energy-efficient because listening con-
sumes energy as well as sending.
2.12. PORT
In [18], authors provide in-network dynamic rate-control
and congestion-avoidance transport scheme. PORT mi-
nimizes energy consumed by avoiding high communica-
tion cost. PORT minimizes energy consumption with
two schemes.
The first is based on the application-based optimiza-
tion approach of sink that feedbacks the optimal report-
ing rates for source nodes. These source report feedbacks
allow the sink to adjust the reporting rate of each data
source. PORT adds a price form each node. Node price is
the total number of transmission attempts made before a
successful packet is delivered from the source and the
sink. It is a metric used to evaluate the energy cost
communication. The sink adjusts the reporting rate of
each source based on the source’s node price and the
information provided about the physical phenomenon.
The second scheme is based on feedbacks from source
node to the sink to inform it about congestion and in-
crease the nodes costs. The sink uses the communication
cost information to slow down the reporting rate of the
appropriate sou rce and increase the repor ting rate of oth-
er sources that have lower communication cost since
reliability must be maintained.
2.13. Flush
Flush [19] is a reliable single-flow bulk transport proto-
col for large diameter WSNs. However, Flush only sup-
ports one data flow. Flush uses an end-to-end reliable
transport protocol to robust to node failures. Flush re-
quires that the sink node sends the sequence numbers of
packets it did not receive back to the data source.
When a source node receives a NACK packet, it re-
transmits the missing data. Flush proposes also a rate
allocation scheme for adapting dynamically the sending
rate of the sensor nodes. This scheme take into count the
broadcast nature of the medium and the interference be-
tween nodes. The rate allocation algorithm follows two
basic rules:
1) Rule 1: A node should only transmit when its suc-
cessor is free from interference.

A. AYADI
111
2) Rule 2: A node’s sending rate cannot exceed the
sending rate of its successor.
These two rules reduce contention and thus collision
in the wireless network and minimize losses due to the
queue overflo ws fo r all no des.
3. Transport Protocol Comparison
In this section, we present a comparison between the
presented protocols in Section 3. The listed protocol can
be classified in categories using some criteria. We can
distinguish between IP-based solution like DTC, TSS
and non IP-based solution like RCRT, ERTP, etc. We
can differentiate between transport protocols by the
manner it recovers losses (end to end recovery or hop by
hop recovery), the use or not have caching in interme-
diate node, kind of message used to loss detection and
recovery (ACK, NACK, IACK) and the level of reliabil-
ity.
3.1. IP and Not IP-Based Solutions
Table 1 presents a classification of all protocol based on
its network and transport layer. DTC and TSS are new
protocols based on IP networks. These two protocols add
caching and hop-by-hop recovery of TCP lost segments.
The use of I P-b as ed tr an sp or t pr o to col enables to con ne ct
between wireless sensor network and an external net-
work (e.g., Internet). Using IP solution in WSN allows
not providing a new proxy at the network border.
3.2. Reliability and Loss Recovery
Reliability is very important for some applications like
health and military applications. Other applications are
loss-tolerant and require a ratio/level of reliability. DTC,
TSS, PSFQ, RMST, RCRT provide 100% reliability for
wireless sensor network applications. However, these
protocols do not use the same mechanism for loss detec-
tion and recovery. For example, ESRT, ERTP, DTSN
and SCTP provide classes of probability for the applica-
tions. However, the need for reliability is not the same
for all applications but depends on the importance of the
application and even on the importance on some packets
than others.
Table 1. IP and not IP-based transport protocol.
TCP/IP Non-TCP/IP
Transport Protocol DTC, TSS RSMT, PSFQ,
ESRT, DTSN,
RCRT, ERTP,
IFRC
Loss detection and recovery methods differ from a
protocol to another. In SCTP and RCRT, the receiver use
gaps in the sequence numbers of received segments as a
signal of packet losses. The receiver asks the retransmis-
sion of missing segments from the source node. This
mechanism is called an end-to-end recovery. However,
in multi-hop wireless sensor networks, end-to-end re-
covery is not energy-efficient and thus new transport
protocols enable intermediate nodes to cache segments.
Loss detection and recovery mechanisms from inter-
mediate nodes allow reducing the total exchanged mes-
sages from the source to the receiver. We distinguish
between two kinds of intermediate nodes. The first does
not detect losses but reacts when it receives a NACK
message. Then, it retransmits the missing segment. The
second kind detects losses and requests a retransmission
from its neighbors.
Acknowledging a packet can be done explicitly by
sending an ACK message, thus if the receiver does not
receive an ACK message, it detects a loss. The explicit
acknowledging approach requires adding a lot of control
messages (ACK) to data messages and increases the
contention. Another solu tion is to relay by a NACK mes-
sage when a packet loss found.
The third mechanism proposed by TSS and ERTP is to
use implicit acknowledgment (IACK). This mechanism
requires that each node i after sending a packet the
next node to the sink overheads the next forwarding. The
forward of a packet by node i is considered as an im-
plicit acknowledgment to node i.
3.3. Congestion Control
Congestion has a significant impact on the performance
of a reliable transport protocol. Some transport protocols
for wireless sensor network (like PSFQ, DTSN, ERTP)
make an assumption that congestion is not likely to be
problem in WSNs. Others assume all packet loss due to
congestion (like DTC, TSS and ESRT).
Some of the proposed solution are distributed, thus all
sensor nodes detect congestion and then share the infor-
mation using a flag named congestion notification.
Thus, the based station, after receiving a message,
which the congestion notification bit is set, slow down
the transmission rate.
On the other hand, centralized solution implements
congestion detection component on the base station. Be-
cause DTC and TSS are extension for TCP of wireless
sensor network, they inherit congestion detection and
avoidance mechanism from TCP. RCRT congestion de-
tection scheme is based on time to recover loss. They
assume that the network is not congested as long as
end-to-end losses are recovered “quickly enough”. Thus,

A. AYADI
112
RCRT permit the sender to transmit at a light rate even if
there are occasional end-to-end losses, since those rate
can be recovered “ qui ckly”.
All other protocols are based on buffer overflows to
signal congestion. SCTP specifies two thresholds: lower
t
and higher t. When the buffer reaches lower t, the conges-
tion bit is set with certain probability. When the buffer
reaches higher t, the node sets the congestion notification
bit in every packet it forwards. ESRT node calculates
1kk
bb b

 , where k b and 1k b are the buffer size
at the end of the
k
th and the

1th k internal. If the
k
bbB  (where B is the buffer size) then the node
signals a congestion by setting the algorithm. IFRC uses
an exponentially weighted moving average of instanta-
neous queue length as a measure of congestion

1
qqqqq
avgwavgw inst . Unlike SCTP, IFRC
uses multiple thresholds

Uk
3.4. Energy Efficiency
Transport protocols should provide reliability with the
minimum number of exchanged messages. This con-
straint comes from the low capacity of energy of sensor
node batteries. ESRT proposes to reduce the sending
events frequency to reduce the total consumed energy.
Others propose to use hop-by-hop recovery instead of
end-to-end recovery to reduce the retransmissions. For
example, RMST, PSFQ, Wisden, DTC and TSS reduce
the amount of exchanged messages by caching not al-
ready acknowledged segments in intermediate nodes and
process a recovery once a lost is detected.
DSTN proposes to reduce the consumed energy by
using Selective Acknowledgment (ACK and NACK)
after an Acknowledgment Windows of messages. DSTN
reduces the number of control messages, which make it
more energy-efficient than other protocols. TSS and
ERTP propose not to use explicit acknowledgment in-
stead of implicit acknowledgment.
These approaches need a cross-layer mechanism be-
tween the link layer and transport layer (e.g. DTC and
TSS). These mechanisms permit to reduce the transport
acknowledgments. However, some of these schemes are
difficult to be implemented in memory-constraint wire-
less devices.
Finally, ART increases the lifetime of the network by
choosing node with more energy capacity to relay pack-
ets from sources to the sink nodes.
All these works have tried to reduce the amount of
control messages in the WSNs and thus increase the life-
time of all the networks. Table 2 summarizes the details
of all the protocols mentioned above.
4. Conclusion
In this paper, we have reviewed recent researches on
reliability and congestion control in wireless sensor net-
works. We presented different methods of packet loss
detection and recovery, congestion detection and avoid-
ance and energy-efficiency. We compared protocol in
term of reliability, congestion control and energy efficiency.
Table 2. Reliable transport protocols.
Transport
Protocol Direction Type of flo ws Congestion Control Congestion Detection End-to-End of Hop-by-Hop Caching Acknowledging
Energy
Aware
RSMT [4] Sensor to Sink Continuous - - Hop-by-Hop
End-to-End Yes NACK/ACK No
PSFQ [2] Sensor to Sink
Sink to Sensor Event-Driven - - Hop-by-Hop Yes NACK No
ESRT [7] Sensor t o Sin k Continuous Yes Buffer size End-to-End No - No
DTC [8] Sink to Sensor - Yes Yes Hop-by-Hop Yes ACK/SACK Yes
TSS [12] Sink to Sensor - Yes Yes Hop-by-Hop Yes IACK/
SACK Yes
ART [13] Sensor to Sink
Sink to Sensor Continuous Event-Driven - - End-to-End No NACK/ACK No
PORT [18] Sink to Sensor Event-Driven Yes Packet loss - - - Yes
Flush [19] Sensor to Sink Continuous - No End-to-End No NACK No
DTSN [14] Sensor to Sink - - Hop-by-Hop Yes ACK/NACK Y es
Wisden [15] Sensor to Sink Continuous - - Hop-by-Hop
End-to-End Yes ACK/NACK Yes
IFRC [16] Sensor to Sink Continuous Yes Buffer size - No - No
RCRT [3] Sensor to Sink Continuous Yes Time to recover
losses End-to-End No NACK No
ERTP [17] Sensor to Sink Continuous - - Hop-by-Hop - IACK Yes

A. AYADI
113
5. Acknowledgements
The work of Ahmed Ayadi has been funded by the Pôle
de Recherche Avancée en Communications (PraCOM).
6. References
[1] J. Yick, B. Mukherjee and D. Ghosal, “Wireless Sensor
Network: Survey,” Computer Networks, Vol. 52, No. 12,
2008, pp. 2292-2330. doi:10.1016/j.comnet.2008.04.002
[2] C. Wan, A. Campbell, and L. Krishnamurthy, “Pump-
Slowly, Fetch Quickly (PSFQ): A Reliable Transport
Protocol for Sensor Networks,” IEEE Journal on Selected
Areas in Communications, Vol. 23, No. 4, April 2005, pp.
862-872. doi:10.1109/JSAC.2005.843554
[3] J. Paek and R. Govindan, “RCRT: rate-Controlled Relia-
ble Transport for Wireless Sensor Networks,” Proceed-
ings of the 5th International Conference on Embedded
Networked Sensor Systems, New York, 2007, pp. 305-
319. doi:10.1145/1322263.1322293
[4] F. Stann and J. Heidemann, “RMST: Reliable Data
Transport in Sensor Networks,” Proceedings of the First
IEEE International Workshop on Sensor Network Proto-
cols and Applications, May 2003, pp. 102-112.
doi:10.1109/SNPA.2003.1203361
[5] C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heide-
mann and F. Silva, “Directed Diffusion for Wireless
Sensor Networking,” IEEE/ACM Transactions on Net-
working, Vol. 11, No. 1, 2003, pp. 2-16.
doi:10.1109/TNET.2002.808417
[6] G. Fairhurst and L. Wood, “Advice to Link Designers on
Link Automatic Repeat ReQuest (ARQ),” IETF, RFC
3366, August 2002.
[7] Y. Sankarasubramaniam, O. B. Akan and I. F. Akyildiz,
“ESRT: Event-to-Sink Reliable Transport in Wireless
Sensor Networks,” Proceedings of the 4th ACM Interna-
tional Symposium on Mobile Ad Hoc Networking &
Computing, New York, 2003, pp. 177-188.
doi:10.1145/778415.778437
[8] A. Dunkels, J. Alonso and T. Voigt, “Distributed TCP
Caching for Wireless Sensor Networks,” Proceedings of
the 3rd Annual Mediterranean Ad-Hoc Networks Work-
shop, 2004.
[9] J. Postel, “Transmission Control Protocol,” IETF RFC
793, September 1981.
[10] H. Balakrishnan, S. Seshan, E. Amir and R. H. Katz,
“Improving TCP/IP Performance over Wireless Net-
works,” Proceedings of the 1st Annual International
Conference on Mobile Computing and Networking, New
York, 1995, pp. 2-11. doi:10.1145/215530.215544
[11] S. Floyd, J. Mahdavi, M. Mathis and M. Podolsky, “An
Extension to the Selective Acknowledgement (SACK)
Option for TCP,” RFC 2883, 2000.
[12] T. Braun, T. Voigt and A. Dunkels, “TCP Support for
Sensor Networks,” Proceedings of Fourth Annual Con-
ference on Wireless on Demand Network Systems and
Services, January 2007, pp. 162-169.
doi:10.1109/WONS.2007.340494
[13] N. Tezcan and W. Wang, “ART: An Asymmetric and
Reliable Transport Mechanism for Wireless Sensor Net-
works,” International Journal of Sensors Networks, Vol.
2, No. 3-4, 2007, pp. 188-200.
doi:10.1504/IJSNET.2007.013199
[14] B. Marchi, A. Grilo and M. Nunes, “DTSN: Distributed
Transport for Sensor Networks,” Proceedings of 12th
IEEE Symposium on Computers and Communications,
July 2007, pp. 165-172. doi:10.1109/ISCC.2007.4381601
[15] N. Xu, S. Rangwala, K. K. Chintalapudi, D. Ganesan, A.
Broad, R. Govindan and D. Estrin, “A Wireless Sensor
Network for Structural Monitoring,” Proceedings of the
2nd International Conference on Embedded Networked
Sensor Systems, New York, 2004, pp. 13-24.
doi:10.1145/1031495.1031498
[16] S. Rangwala, R. Gummadi, R. Govindan and K. Psounis,
“Interference Aware Fair Rate Control in Wireless Sensor
Networks,” Proceedings of the Conference on Applica-
tions, Technologies, Architectures, and Protocols for
Computer Communications, New York, 2006, pp. 63-74.
[17] T. Le, W. Hu, P. Corke and S. Jha, “ERTP: Ener-
gy-Efficient and Reliable Transport Protocol for Data
Streaming in Wireless Sensor Networks,” Computer
Communications, Vol. 32, No. 7-10, 2009, pp. 1154-1171.
doi:10.1016/j.comcom.2008.12.045
[18] Y. Zhou, M. Lyu, J. Liu and H. Wang, “PORT: A
Price-Oriented Reliable Transport Protocol for Wireless
Sensor Networks,” Proceedings of 16th IEEE Interna-
tional Symposium on Software Reliability Engineering,
November 2005, pp. 10-126. doi:10.1109/ISSRE.2005.32
[19] S. Kim, R. Fonseca, P. Dutta, A. Tavakoli, D. Culler, P.
Levis, S. Shenker and I. Stoica, “Flush: A Reliable Bulk
Transport Protocol for Multihop Wireless Networks,”
Proceedings of the 5th International Conference on Em-
bedded Networked Sensor Systems, New York, 2007, pp.
351-365. doi:10.1145/1322263.1322296