MoMoRo: Providing Mobility Support for
Low-Power Wireless Applications
JeongGil Ko and Marcus Chang
Abstract—Recently, mobile devices have been introduced in various
wireless sensor network (WSN) applications in order to solve complex
tasks or to increase the data collection efficiency. However, the
current generation of low-power WSN protocols is mainly designed to
support data collection and address appli-cation-specific challenges
without any particular considerations for mobility. In this paper, we
introduce MoMoRo, a mobility support layer that can be easily
applied to existing data collec-tion protocols, thereby enabling
mobility support in the network. MoMoRo robustly collects
neighborhood information and uses a fuzzy estimator to make link
quality estimations. This fuzzy estimator continuously reconfigures its
thresholds for determining the fuzzy sets, allowing MoMoRo to easily
adapt to changing channel environments. Furthermore, MoMoRo
includes an active destination search scheme that allows disconnected
mobile nodes with sparse traffic to quickly reconnect if there are
packets in the network destined to this mobile node. We evaluate
MoMoRo both indoor and outdoor and show that a continuously
moving device in a MoMoRo-enabled RPL (i.e., IPv6 Routing Protocol
for Low-Power and Lossy Networks) network can achieve a high
packet reception ratio of up to 96% and stay connected in areas where
RPL alone cannot and with less than half the packet overhead needed
by the well-known Ad hoc On-Demand Distance Vector routing
protocol.
Index Terms—Fuzzy estimation algorithm, Internet Engineer-
ing Task Force Routing Protocol for Low-Power and Lossy Net-
works (RPL), node mobility support.
I. INTRODUCTION
W ITH the emerging cyber physical systems (CPS) and Internet of
Things (IoT), low-power wireless networks are now taking their next step
toward widespread usage in
various everyday life applications. Examples of such applica-tions
include wireless physiological data sensing for clinical care,
military-related sensing applications, and robotic network systems
for various automation and sensing tasks [4], [5], [20],
[17]. A decade worth of research by the wireless sensor net-
work (WSN) community provides the foundation for forming
such wireless networks with proven network layer protocols
that addresses the unique challenges in low-power and lossy
Manuscript received August 10, 2012; revised September 5, 2013 and
November 5, 2013; accepted December 20, 2013. This work was supported by
the Korea Evaluation Institute of Industrial Technology and the Korean
Ministry of Science, ICT and Future Planning under Project 10041725 (De-
velopment of application service and software to support sensor terminals for
providing personalized service based on smart devices).
J. Ko is with Electronics and Telecommunications Research Institute,
Daejeon 305-350, Korea (e-mail: jeonggil.ko@etri.re.kr).
M. Chang is with the Department of Computer Science, Johns Hopkins
University, Baltimore, MD 21218 USA.
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JSYST.2014.2299592
networks (LLNs). Protocols, such as Collection Tree Protocol
(CTP) [11] and Routing Protocol for Low-Power and Lossy
Networks (RPL) [30], allow a system developer to simply choose
a network layer protocol that suits the target application well rather
than designing a completely new routing protocol.
However, in these new CPS and IoT applications, mobil-ity
and actuation becomes an integral part of the low-power
wireless network and cause radically different network traffic
patterns than those found in traditional data collection appli-
cations. Hence, protocols specifically designed for stationary
networks are ill equipped to tackle the low-power data collec-
tion challenges found in hybrid networks with both stationary
and mobile nodes. Furthermore, they cannot provide adequate
performance in these applications; therefore, a new approach is
needed to address the mobility issue in these networks.
Previous efforts to support mobility in LLNs show that a
multiple-tier network architecture that separates the role of data
forwarding from the mobile nodes is effective in achieving high
reliability, but such an architecture requires a large set of nodes
to form the backbone network and require careful consideration
in the deployment phase [7], [16].
While this two-tier architecture captures the essence of a hybrid
network, we seek a solution with less rigid division between data
forwarding and mobile nodes, thereby reduc-ing the constraints
imposed on the underlying data collection backbone. Furthermore,
we target low-power wireless network applications in which 1)
human-scale mobility exists, 2) only low amount of traffic is
transmitted in the network, and 3) data collection is the main traffic
during the majority of the time, but point-to-point traffic is used to
deliver urgent alerts or configuration/actuation messages to
individual nodes. Given the maturity of the available low-power
wireless network proto-cols and that data collection is still a
cornerstone CPS and IoT application, we draw from previous
research efforts into data collection protocols and leverage the
routing capabilities these protocols provide. Specifically, we
introduce a mobility support layer, called MoMoRo, to ensure that
high packet delivery rates can be achieved even with mobile nodes
in the network. At the same time, MoMoRo does not introduce
significant performance overhead when all the nodes are
stationary. Our previous work shows that the reason behind this
performance degradation with current generation collection
protocols is the rapid change in the wireless environment caused
by nodes moving into areas with different wireless characteristics.
In other words, protocols for stationary networks are designed to
compensate for temporal changes in the wireless environment,
whereas the spatial changes caused by mobility happen too rapidly
for these protocols to adapt in time [16].
 Hence, in order to effectively support device mobility in LLNs,
the mobility support layer must be able to provide robust link
quality estimations for the changing neighborhood. In addi-tion,
since bidirectional communication is an important feature in CPS
and IoT applications, the mobility support layer must also be able
to assist the underlying routing protocol to quickly reestablish its
bidirectional routes. MoMoRo accomplishes this by actively
transmitting probes once it detects that the node has been
disconnected from its current parent in the data collection tree.
Link quality estimators, such as the expected number of
transmissions (ETX), received signal strength (RSS), and symbol
error rate (SER) variance, collected from the probes are analyzed
using a fuzzy logic estimator to determine good quality links in the
new neighborhood. Using a fuzzy estimator allows us to
compensate for the uncertainties in the wireless channel conditions
while effectively classifying the wireless links’ quality.
Furthermore, to allow the fuzzy estimator to easily adjust to
varying environment conditions, we use an adaptive scheme to
continuously reconfigure the thresholds used to determine the
fuzzy sets. Links that are estimated to have good quality by the
fuzzy estimator are suggested to the underlying routing protocol so
that it can select a route that optimizes its respective end-to-end
metric. Additionally, MoMoRo includes an active destination
searching scheme that initiates adaptively scoped floods to search
for “missing” destinations. This scheme is activated when a mobile
node has not yet realized that it has been disconnected from the
previous route; thus, a route reconfiguration has not been initiated
at the new neighborhood, and the node is missing from its previous
neighbors.
Our contribution in this paper is the mobility support layer
MoMoRo, which we evaluate using a series of experiments
with stationary and mobile Tmote Sky devices [23] installed in
an indoor testbed and validate the performance in an outdoor
environment as well. Our results show that MoMoRo supports
the underlying collection protocols well when human-scale mo-
bility is introduced, with high packet reception performance and
low additional packet overhead. We also present performance
comparisons with the Ad hoc On-Demand Distance Vector
(AODV) routing protocol, which is specifically designed for
mobile networks. Our results indicate that MoMoRo achieves
up to 96% packet delivery performance with less than half of
the overhead introduced by an AODV implementation designed
specifically for LLNs.
The structure of this paper is as follows. Section II introduces
related work in three different related research domains. Next, in
Section III, we present MoMoRo as a mobility support layer for
low-power wireless networks and evaluate its performance in
Section IV. Finally, we conclude this paper with a summary and a
list of future research direction in Section V.
II. RELATED WORK
Research on supporting mobile devices in wireless net-works
has been widely studied in various wireless technique domains.
We use this section to present an overview of the existing
literature in various domains and position our work
accordingly.
A. Mobility in LLNs
Numerous efforts exist to provide a routing protocol with
high packet reception ratio (PRR) for WSNs. One of the most
successful efforts is the CTP [11]. As the default data collection
protocol in TinyOS 2.x, CTP provides high PRR routing paths
with minimal end-to-end costs by provisioning routes based on
the four-bit link estimator [10]. While CTP performs well for
stationary node data collection, it lacks the support for routing
to individual nodes in the network and was not designed to
provide any form of mobility support. More recently, Hydro
and RPL were introduced to provide IPv6 routing to LLN
systems. These protocols provide support for data collection
traffic as well as point-to-point traffic [9], [30]. However, like
CTP, Hydro and RPL were not designed to support mobile
nodes. Further mobility-related work in sensor networks either
only support sink node mobility [21], require local or global
time slots [3], or require the use of global location infor-mation
[1].
WSN systems such as MEDiSN [16], CodeBlue [22], and the
Washington University physiological data collection system [7]
have proposed some customized solutions to support mobile
devices in a WSN. MEDiSN and the Washington University
physiological data collection system form a two-tier network by
simplifying the routing capabilities of the mobile devices and
placing a designated backbone routing tree network of stationary
devices for data forwarding. While such an approach simplifies the
issue of maintaining continuous connectivity to mobile devices,
one drawback is that, in order to increase the connectivity range,
the backbone needs to be extended, even when some locations of
the coverage area may not have any traffic to forward (e.g., an area
with only infrequent node visits needs the same backbone
infrastructure as a well-visited area). In addition, to ensure high-
quality links, the deployment of the backbone network requires
careful provisioning [7], [16]. The CodeBlue system incorporates
a fully ad hoc routing protocol and can theoretically solve this issue
[6]. However, the performance of CodeBlue quickly degrades with
increasing hop counts, large number of devices, and user mobility
[22]. Our approach is an effort to combine these two approaches
and tries to maintain a routing tree, which includes both stationary
and mobile devices, while effectively supporting connectivity for
collection and point-to-point traffic with mobile nodes on top of
this base routing architecture.
B. MANETs
The well-known AODV routing algorithm is a reactive rout-ing
protocol designed for mobile ad hoc networks (MANETs)
[24]. While AODV adapts to networks with frequent link
changes, the high overhead and delay from the route discovery and
maintenance makes it less suitable for resource-limited networks
such as LLNs. Nevertheless, implementations such as TinyAODV
and Not So Tiny (NST)-AODV represent efforts to bring AODV
to LLN applications [13], [27]. We will use NST-AODV, an
improvement to TinyAODV, in our evaluations to show how an
AODV implementation for LLNs performs with a typical data
collection application and human mobility.
 Dynamic source routing (DSR) is another on-demand routing
protocol for MANETs that uses source routing for packet
forwarding [15]. By using source routing for all the packets, it
introduces a relatively high overhead. In addition, because a node
must have knowledge of the entire route before attaching the
source routing header, and the fact that DSR does not have a local
route repair mechanism for broken links, the overhead of
maintaining the routes can become very expensive. Finally, in
DSR, path maintenance (e.g., new route requests) occurs when an
already transmitted packet encounters a broken link; thus, this
recovery phase can introduce a large latency.
Furthermore, most MANET routing protocols aim to estab-
lish end-to-end paths for data transmissions [2]. In other words,
when a single link on the path breaks (e.g., due to mobility or
link fluctuations), the entire end-to-end path is reconstructed,
increasing routing overhead and latency. Similar overheads are
also encountered when using proactive routing schemes such as
Destination-Sequenced Distance Vector routing [25]. On the
other hand, our scheme’s basic routing paths are based on the
paths selected by a tree-based routing protocol, and we focus
on maintaining these paths using the assistance from our
mobility support layer to perform local routing path repairs
with minimal overhead.
Finally, we observe that mobile LLNs carry different traffic
types than typical MANETs. Typically, in MANET research,
the main focus of the routing protocols is to establish any-to-
any routing paths; whereas, for mobile LLNs, collection is the
main traffic pattern, and point-to-point traffic support is an
application-specific requirement and deals with a significantly
smaller number of packets [30]. It is therefore reasonable for
mobile LLNs to optimize their routes for the collection traffic
and use the already established routes to support additional
traffic patterns.
C. Mobile Device Handoff
Many wireless research domains propose device handoff
schemes to provide seamless connectivity on mobile devices
(see [19], [28], and references in [26]). However, we argue that
for LLNs, such techniques are too aggressive. Given that the
primary consideration in designing an LLN system is to
optimize for the limited resources available on the devices (e.g.,
bandwidth, energy, etc.), wasting packets to perform detailed
analysis of whether a link would break or not or to select the
best possible link at any point in time [8], [14] is a luxury.
Rather, the characteristics of LLNs suggest that we utilize a
scheme to quickly and accurately recover from link losses and
find new routes. In other words, a reactive scheme is more
suitable for LLNs than a proactive scheme in which we try to
predict future link disconnections.
III. MoMoRo: THE MOBILITY SUPPORT LAYER
MoMoRo supports node mobility by performing two major
operations. First, it tries to react quickly to packet losses to either
reconstruct the default route (e.g., route toward the root of the
routing tree) or notify a nonroot destination that its route is
currently unavailable. Second, when reconstructing the default
route, it quickly gathers neighborhood information and passes
Fig. 1. Diagram of how MoMoRo interacts with the network stack.
this information through a fuzzy estimator to identify links that are
currently good and usable to hint the routing layer in making its
route selections. The remainder of this section provides detailed
descriptions of the MoMoRo mobility support layer.
A. Mobility Support Layer in the Networking Stack
As Fig. 1 shows, the operation of MoMoRo is implemented
in the networking stack as an additional layer between the link
layer and the routing layer. We note that only mobility/routing
configuration-related information/packets are exchanged be-
tween the three layers. All other packets, such as typical data
packets, do not pass through the MoMoRo layer and thus do
not add any additional processing overhead. The main benefit
of implementing MoMoRo as an independent layer rather than
a software module within the routing layer is that it potentially
provides the flexibility to easily apply and reuse the MoMoRo
code with different routing protocols rather than reimplement-
ing the procedures when the routing layer changes.
B. Realizing a Link Disconnection
Typically, a routing layer collects information on the number of
transmission attempts for each unicast packet to perform link
quality estimations (see Fig. 1). Once a node realizes that a packet
transmission to node n has failed (i.e., the number of
retransmissions reaches the maximum), the routing layer notifies
MoMoRo that the link with node n may be potentially lost. At this
point, MoMoRo makes one last attempt to reach n by transmitting
a unicast packet. This attempt allows the node to suppress any
additional overhead due to route reconstruction if the reason
behind the failed transmission is due to instantaneous link quality
fluctuations. If an acknowledgment is received, no further action
is made. However, if the acknowledgment is not received in
timetinit_probe, MoMoRo first determines if n is the default route.
If this is true, an operation to reconstruct the default route takes
place (see Section III-C); otherwise, MoMoRo will start the search
for the packet’s final destination since the route to this missing
destination node is broken (see Section III-D).
C. Reconstructing Default Routes
Once a nodeA determines that the default route is lost and,
therefore, the data collection tree needs to be repaired, Mo-
MoRo quickly broadcasts countbroadcast messages to its (new)
wireless neighborhood. Nodes that hear this broadcast will send a
unicast reply. Using the received replies, node A computes the
ETX of the link, the average RSS on the link, and the link’s SER
variations. We explain the three metrics in detail below.
1) ETX: ETX deals with the link quality in an experimental way
and represents the expected number of transmission attempts
needed to make a single successful packet trans-mission
[30]. In wireless networking, ETX is considered as one of
the most accurate link quality estimation metrics
for various routing protocols [30], [11]. In MoMoRo, we
compute the ETX using the countrx replies that we
receive from each link as follows:
count
ETX = broadcast
countrx
where countrx is the actual number of probe packets that
are received.
2) Average RSS: The average RSS of the received packets,
i.e., RSS, can provide us with a rough estimate of the
expected connection quality. We compute RSS only
using countrx packets.
3) SER Variance: Previous work shows that the SER vari-
ance increases noticeably as the distance of the link
increases to a point where potential link disconnections
may occur [18]. On both the popular TI CC2420 and
CC2520 radios, the link quality indicator (LQI) values of
probe packets can provide us with a rough estimate of the
SER encountered as a packet travels a wireless link. Since
all IEEE 802.15.4 radios are required to report an LQI
value, this metric is widely available on WSN platforms.
We compute the variance of the SER (LQI) values using
the following equation:
σSER2 = SER2 − SER2.
We point out that the metrics introduced above have their own
benefits and drawbacks. Specifically, while the ETX metric is an
accurate measure of link quality, it takes multiple packets to
achieve reasonable accuracy. Furthermore, the RSS metric can be
collected quickly but can show high variance with changes in the
channel conditions, and lastly, σSER2 presents an accurate
evaluation of a link condition with minimal latency but typically
shows the link quality as a binary metric (e.g., good or bad) [18].
To compensate for these benefits and drawbacks, we need a way to
intelligently combine the metrics so that we can exploit the benefits
of each metric. Furthermore, for each metric, we should come up
with a single answer to the following question: “Is this neighbor
good enough to connect to?” The simplest and possibly the most
accurate answer to this question could be obtained by normalizing
each metric, identifying the relationships between various metrics
and then using an optimization algorithm to end up with a
probability of potential link (dis)connection. However, in a
dynamic channel environment, it is difficult to get an accurate
relationship func-tion over different link metrics, and the resulting
probability would require another threshold for the algorithm to
output a firm decision (i.e., yes or no).
As an alternative, we consider the use of a fuzzy-logic-based
link quality estimator. Using fuzzy logic is beneficial
Fig. 2. Diagram of our fuzzy membership function for determining the fuzzy
2
sets ofETX, RSS, and σSER . The values of the Ax variables are application
and channel environment dependent or can be adaptively set.
given that it provides a structured environment for multiple input
variables to result in a single output. In addition, since it is difficult
to extract the meaning of a metric using static thresholds in a
dynamic channel environment, by using a fuzzy-logic-based
estimator, we can take the advantage that the input metric does not
need to be exact but rather can cope with some unexpected
variations using overlapping regions in defining the fuzzy sets.
Specifically, each metric value is passed through a fuzzy
membership function, where the fuzzy set is determined for each
metric (e.g., input metric value is x% “good,” y% “medium,” and
z% “bad”). By summing up the result of the fuzzy membership
function for each metric, we can predict the expected quality of a
link in various dimensions. The output of the fuzzy estimator will
indicate if the link that is being diagnosed is expected to be either
“good,” “ok,” or “bad.”
The main challenge in defining an effective fuzzy estimator is
properly defining the fuzzy sets and the membership functions. For
this, we first start by defining a simplified and general diagram of
a fuzzy estimator, as shown in Fig. 2. In other words, defining the
variables A0, A1, A2, and A3 for each input metric (e.g., ETX,
RSS, and σSER2) becomes a critical issue since configuring
inefficient set boundaries will naturally lead to improper link
quality detection. Using the base diagram in Fig. 2, the rest of this
section discusses about how we configure the fuzzy sets for each
input metric that we target.
1) ETX: Among the input metrics for our fuzzy estimator,
the fuzzy sets for the ETX metric can be easily config-
ured to meet application-specific targets. For example, an
application may consider links with a 90% PRR as good
and therefore set the threshold for ETX’s fuzzy
membership function as 1.11 Tx. Furthermore, if 50% is
the threshold for determining bad links, this can be
translated as 2.00 ETX and used as the threshold for
ETX’s fuzzy membership function. These thresholds can
be set to tolerate for some unexpected variations (e.g.,
5%), thereby determining A0 = 1.05, A1 = 1.18, A2 =
1.82, and A3 = 2.22.
2) σSER2: For the fuzzy membership function of σSER2, our
measurements and previous work suggest that 0 < σSER2
<∼ 3 indicates a good link (e.g., link with a short distance
and a high expected PRR), ∼ 3 < σSER2 <∼ 50 indicates
a link that works well with some fluctuations, and σSER2
>∼ 50 indicates a link with bad quality [18]. Again, since
we are defining fuzzy thresholds, we can further define
safety bounds (i.e., ranges of A1−A0 and A3−A2) for the
σSER2 metric to account for realistic variations and
uncertainties. As a result, for σSER2, one

Fig. 3. Flowchart of the operations in MoMoRo and an illustration of how
the threshold values for RSS’s fuzzy membership function is dynamically
configured. In the flowchart, t indicates tlink_validation, and CountG indicates
count . that the known route is no longer valid. The same can happen when
good_limit
can define the fuzzy membership function using A0 = 1,
A1 = 5, A3 = 45, and A4 = 55.
3) RSS: On the other hand, given the dynamics of the RSS
values, it is difficult to determine static threshold values that
work well in various environments for the RSS metric
[18]. For this reason, we introduce a simple adaptive
threshold configuration scheme that helps estimate proper
values for RSS’s threshold values in the fuzzy member-ship
function. We assume in this scheme that the safety bounds
(i.e., regions A1−A0 and A3−A2) are fixed to a reasonable
range (e.g., 10 dBm), and the two thresholds that define a
bad RSS are also fixed to a reasonable low value (e.g., A2 =
−84 dBm and A3 = −94 dBm).
As Fig. 3 illustrates, once a link is confirmed to be discon-
nected, MoMoRo checks how long this link has been used as the
default route. If this duration is shorter than tlink_validation, both
A0 and A1 values are increased by 5 dBm (a different increase
value can be used). This decision is made from the ob-servation
that a quick disconnection of the selected link implies that the
output of the fuzzy estimator did not successfully output a good
enough link. Therefore, MoMoRo should increase the
bar for selecting a good link and try to assure that links that fall in
the “good” category have better quality links than before.
Upon evaluating the potential links in the fuzzy estimator for
RSS, if the number of links that are classified as good is too small
(e.g., < countgood_limit), MoMoRo will try to decrease the values
of A0 and A1 by 5 dBm and recursively run the fuzzy estimator
with the same links once more. This processes repeats until more
than countgood_limit are classified as good or until A1 reaches A2.
This decrease in thresholds is to assure that MoMoRo can supply
the routing layer with enough links to compare from and to make
sure that the increase in A0 and A1 values from the previous case
does not filter out all of the potential links by setting a bar that is
too high. In practice, and in our evaluations, countgood_limit is set
to 1.
Given that RSS is a metric that is heavily dependent on the
channel conditions, this adaptiveness of the fuzzy estimator
allows MoMoRo to be easily deployed in various environments
in which the wireless channel dynamics are difficult to model
or even unknown. We will show in our evaluations that using
this adaptiveness helps reduce the amount of overhead needed
to identify good links.
D. Actively Searching for Destinations
When a mobile node is actively transmitting data (either its own
or for others), the node can immediately detect the loss of its
default route by realizing that its packets are not getting through
and initiate a reconstruction as described in the previous section.
Therefore, packets that are headed to this node (e.g., point-to-point
traffic), can be forwarded to the newly reconstructed route.
However, if a mobile node is not transmit-ting or receiving
packets, it might have moved to a different physical location in the
network without any nodes (including itself) being aware of it.
However, when some other nodes try to send the mobile node any
packets, a node on the route toward the mobile node will discover
the number of packets gen-erated is low in the network, which is a
common case for many sensor network and LLN applications. To
address such cases, MoMoRo includes an active destination
searching scheme.
This scheme is activated when the routing layer reports that
the connection to the next hop node, which is not the default
route, is potentially missing. At this point, the node will issue
an adaptively scoped flood to find the final destination of the
packet. First, the node will ping the nodes in the s = scopeinit hop
neighborhood to check if the final destination is within the
s hop scope. scopeinit is set to 2 given that, if the destination was
in the one hop neighborhood, the active destination search would
have not begun. If a node listens to this flood packet and it is not
the destination, the node will stamp its address to the packet and
make another broadcast as long as it has not exceeded the scope
limit of the packet. Once the packet reaches the final destination, a
reply is sent to the origin of the flood using the intermediate nodes’
addresses in the flood packet as a source routing header. If the
acknowledgement of the flood is not received at the origin within
a time interval, the value of s is incremented by one, and the flood
is reinitiated. In addition to generating the reply packet, once the
destination node receives

Fig. 4. Node positioning and mobility trace of our indoor testbed environ-ment.
The collector node is positioned at one end of the building, and we present the
trace of mobility in dotted lines.
the search flood packet, it realizes that its default route is no longer
valid and actively starts to send probe messages to learn about its
neighborhood and reestablish its routes using the opera-tions in
Section III-C to reconstruct its bidirectional default route.
IV. EVALUATION
We evaluate the performance of MoMoRo in an indoor testbed
environment consisting of 30 Tmote Sky devices [23] distributed
over the second floor of an office building (see Fig. 4). The Tmote
Sky includes the TI/Chipcon CC2420 IEEE 802.15.4 radio [29],
which is a low-power radio platform, and our TinyOS
implementations offer the capability to collect RSS and SER
readings based on the first eight symbols of each received data
packet. We evaluate our system using three different performance
metrics. First, we compute the PRR at the final destination to
evaluate the application level efficiency of MoMoRo. Second, we
compute the end-to-end number of transmissions, which represents
how many transmissions (in-cluding retransmissions) a packet
experienced from the source to the destination. This metric
represents the efficiency and ef-fectiveness of the selected routes.
Finally, we count the number of overhead packets issued per node
per minute to monitor the cost of establishing quality routes. We
present results for MoMoRo implemented together with the RPL
routing protocol using the Minimum Rank with Hysteresis
Objective Function (MRHOF) [12]. We also have an
implementation of MoMoRo with CTP, but we use RPL in our
evaluations since RPL pro-vides support for more types of traffic
patterns. For simplicity, we call this network as MoMoRo and
compare MoMoRo against the standard-compliant RPL with
MRHOF implementa-tion and with a MANET routing protocol,
NST-AODV (simpli-fied as AODV throughout this section). All
implementations are in TinyOS 2.x. For MoMoRo’s fuzzy
membership functions, we use A0, A1, A2, and A3 values, as
specified in the previous section. We note that the channel
environments for all our experiments were done on Zigbee/IEEE
802.15.4 channel 26 (e.g., no WiFi interference) and in a stable
environment with
minimal external interfering factors (e.g., human movement,
environment changes, etc.).
We begin our evaluations with the simplest but most impor-
tant network configuration for typical LLN systems: a station-
ary network where a single node collects data from all the other
nodes, in order to measure the overhead caused by MoMoRo in
a network without mobility.
Unless specified, all our results are the average of five runs taken
from different times of the day. All source nodes generate data
packets at an interval of 5 s (e.g., 0.2 Hz). We point out that we
observed similar results when configuring this interval differently
(e.g., 0.1 and 1 Hz). We unify the parameters that affect
performance such as the number of maximum link-layer
retransmissions, the retransmission interval, the message queue
size, and the routing table timeouts so that all three protocols use
the same values. The data collection node was configured to be at
one side of the testbed (see Fig. 4), and 40% of the testbed was
reachable (i.e., PRR > 50%) in one hop with a radio trans-mission
power of −3 dBm, and the node farthest away could be reached in
four hops. We did not further decrease the radio transmission
power in order to ensure the testbed’s connectivity.
A. Stationary Network With Collection Traffic
Fig. 5(a) presents the end-to-end PRR for networks running
MoMoRo, RPL, and AODV. The results show that all three
protocols are able to achieve very high PRRs in a stationary
collection network. Fig. 5(b) shows that RPL achieves the high-
est PRR with the smallest amount of overhead when compared
with AODV and MoMoRo, which is not surprising since this is
the exact scenario RPL was designed for.
One surprising thing to notice, though, is that the overhead
introduced by AODV is not significantly different from the other
two protocols. In one of our experiments conducted at night (i.e.,
minimal changes to the link qualities over time), the PRR for
AODV was 99.98%, whereas the amount of overhead was only
half of MoMoRo and RPL. This is due to the fact that RPL (itself
and when used with MoMoRo) employs a proactive approach in
choosing the most suitable link, i.e., rout-ing beacons are
continuously transmitted (based on a Trickle timer) and new
(better) routes are chosen even when the old ones are still valid.
AODV, on the other hand, relies on a more reactive approach
where new routes are only considered when the old ones are lost.
Hence, in the nighttime experiment, AODV did not send any
routing beacons, whereas the link that it chose stayed stable. 1
However, in the daytime experiments with AODV, we observed
that the cost of reconfiguring routes once they are lost due to link
quality fluctuations is significantly higher; thus, the values for
AODV presented in Fig. 5(b) show a similar average but with a
larger variation. We see a similar effect of MoMoRo and RPL’s
proactiveness in Fig. 5(c), where we present the average number of
end-to-end transmissions that a successfully received packet was
transmitted in the network. While the PRR and the number of
overhead packets are similar (or even lower in some cases) for
AODV, the routes selected by
1The original proposal of AODV includes periodic beaconing for route maintenance, but this
was not implemented in NST-AODV.
Fig. 5. Results collected from a network consisting only of stationary nodes with collection traffic. (a) PRR for MoMoRo, RPL, and AODV. (b) Per minute
overhead packets generated at each node when running MoMoRo, RPL, and AODV. (c) End-to-end number of transmissions occurring for each successfully
received data packet for MoMoRo, RPL, and AODV.
TABLE I
PRR AND PACKET OVERHEAD PER MINUTE AT EACH NODE FOR
MOMORO WHEN ONLY RSS IS USED AS THE FUZZY ESTIMATOR’S
INPUT. WE TEST FOR THE CASE WITH THE
PROPOSED ADAPTIVE THRESHOLD CONFIGURATION SCHEME AND ALSO TEST
CASES WITH STATIC THRESHOLDS OF −10, −50, AND −80 dBm FOR DETERMINING
GOOD LINKS
MoMoRo and RPL are more efficient, in terms of end-to-end
number of transmissions because these protocols can choose
better links once they are available, whereas AODV has to wait
until the connection has been lost first.
In Table I, we show the PRR and overhead performance of
MoMoRo when only RSS is used as the single metric for the fuzzy
estimator. We test for the cases where we use the proposed
adaptive threshold controlling for the RSS’s fuzzy set and for static
thresholds of −10, −50, and −80 dBm. First, we notice that, when
only using RSS, while we still see a high PRR, the amount of
overhead is higher than the case where all three metrics are used
[see Fig. 5(b)]. In addition, we notice that, by using the adaptive
threshold configuration scheme, we successfully decrease the
amount of overhead in the network by adapting the threshold of
classifying good RSS fuzzy sets to match the current channel
conditions, which can be different for each individual link. In
summary, this shows that the adaptive fuzzy estimator is able to
adapt to unknown or varying wireless signal conditions, thereby
improving the PRR and overhead.
B. Stationary Network With Collection and Downward Traffic
Next, we examine how our three protocols of interest
perform if we have bidirectional traffic. For this, we take the
same net-work configuration as above and configure the data
collection node to also periodically send traffic to all nodes in
the network sequentially with the same packet interval of 5 s.
Fig. 6(a) shows the PRR for the upward collection traffic and
the downward point-to-point traffic. We conjecture that the
PRR discrepancy that we observe between the two types of
network traffic, for both MoMoRo and RPL, is caused by
asymmetric links. Since both MoMoRo and RPL try to use the
reverse of the optimized upward (collection) routes to reach the
downward (point-to-point) destinations, the quality of the
downward links is neither always optimal nor identical to the
upward links. On the other hand, AODV maintains separate
routes for all destinations. Therefore, the PRRs of downward
traffic and collection traffic in AODV are similar.
We notice in Fig. 6(b) and (c) similar trends in overhead and
number of end-to-end transmissions. One thing to point out
here is that the amount of overhead increase is more significant
for AODV than the other two protocols when compared with
Fig. 5(b). This is due to the fact that AODV tries to maintain all
of its routes separately, whereas MoMoRo and RPL use the
routes that are established for collection traffic. Since MoMoRo
reacts quickly to packet losses when dealing with downward
traffic, the amount of overhead increase is higher for MoMoRo
than RPL.
In summary, the preceding results show that adding the
MoMoRo mobility support layer to a collection protocol does not
significantly degrade the performance of the network when the
network is stationary, regardless of the two traffic patterns.
C. Introducing a Mobile Node to the Network
In this experiment, we first add a single mobile node to
examine the effect of mobility on the network performance.
This mobile node transmits a packet toward the data collection
node, and the data collection node sends one packet to this
mobile node every second (e.g., 1 Hz), whereas all the other
nodes in the network send one packet every 5 s (e.g., 0.2 Hz) to
the collection node. We increase the packet transmission
frequency for the mobile node to see the effect of mobility in
more fine grain. The mobile node moves through all parts of
where the stationary testbed nodes are positioned (see dotted
line in Fig. 4) for a duration of 150 s.
Fig. 7 presents the PRR achieved for the collection and
downward traffic from/to the mobile node while it was moving.
Notice that for MoMoRo and AODV, the PRR in all cases is above
90%, whereas the PRR for RPL, which is not designed to support
mobility, significantly degrades. Taking a further look at the
amount of overhead packets per node generated during this mobile
period in Fig. 8, we notice only a minor increase in packet
overhead for RPL [when compared with Fig. 6(b)], meaning that
the protocol did not react properly to the link disconnections
caused by the mobile node. In other words, compared with
MoMoRo and AODV, the neighborhood discovery and link
estimations of RPL were not robust enough
Fig. 6. Results collected from a stationary network with collection and downward traffic. (a) Collection and downward PRR for MoMoRo, RPL, and AODV.
(b) Per minute overhead packets generated at each node when running MoMoRo, RPL, and AODV. (c) End-to-end number of transmissions occurring for each
successfully received data packet for MoMoRo, RPL, and AODV.
Fig. 7. PRR of the collection and downward traffic from/to the mobile node for
MoMoRo, RPL, and AODV.
Fig. 8. Amount of overhead introduced during the period when one node was
mobile in the network.
to react to such unexpected disconnections. On the other hand,
both MoMoRo and AODV start taking more aggressive actions,
leading to a larger overhead but also a high PRR, with the
overhead from the AODV network almost twice as large as the
MoMoRo network.
We note that our experiments were conducted at various times
during the day. Specifically, for each protocol, we run five
experiments, evenly spaced, during 10 A.M. to 1 A.M. As a result,
despite the fact that our experiments were done using a single
testbed, the temporal variations of the channel environment over
the day introduce a new channel environment for the networks to
operate in. The small variations we see with MoMoRo suggest that
it can adapt well to various channel conditions when compared
with AODV or RPL.
In order to examine the packet loss distribution over time, we
select a single trace from each configuration for comparison taken
at nighttime when the wireless channel was stable. Fig. 9 presents
the mobile node’s collection PRR (top) and point-to-
Fig. 9. (Top) Collection and (bottom) downward PRR for a mobile device over
time using 30-s buckets. The first and last two buckets show the PRR before
and after the mobile phase, whereas the five buckets in the middle show the
PRR of MoMoRo, RPL, and AODV during the mobility phase.
point PRR (bottom) for a total of 4.5 min (i.e., 1 min before the
mobility occurred, 2.5 min of the mobility period, and 1 min after
the mobility ceased) split into 30-s bins. We notice that, while
MoMoRo and AODV show a continuous high PRR over the entire
period, RPL’s PRR performance drops as mobility is introduced.
In fact, during the fifth 30-s period, the mobile node was
completely disconnected from the network (resulting in a PRR of
0%). Even during this period, we see in Fig. 10, where we plot the
amount of overhead packets in the entire network during the same
time frame, that RPL did not increase its amount of overhead
packets in order to discover new routes. On the other hand, when
losses occur, both MoMoRo and AODV start generating overhead
packets to reconfigure the routes, and as a result, both achieve high
PRR. In this case, AODV
Fig. 10. Number of overhead packets in the network over time using 30-s
intervals. The first and last two cases show the overhead packets generated
before and after the mobile phase, whereas the five cases in the middle show
the overhead packets generated during the mobile phase.
Fig. 11. PRR and number of overhead packets in the network over time with
three mobile and ten stationary transmitters in a 33-node network. The first and
last two cases show the overhead packets generated before and after the mobile
phase, whereas the five cases in the middle show the overhead packets
generated during the mobile phase.
requires more packets to reconstruct its routes since the route
reconstruction happens in both directions. Fig. 8 also shows the
same trend as well. For example, even if the collection route is
reconstructed, a downward packet will encounter a nonvalid route
and try to reconstruct the downward route separately. On the other
hand, MoMoRo benefits from the fact that the underly-ing routing
layer, RPL, tries to use the reverse of the collection routes as the
downward routes. Therefore, once the default collection route is
reconstructed, packets traveling downward can use the updated
route without additional packet overhead.
Finally, in Fig. 11, we plot the PRR and overhead packet count
of MoMoRo after increasing the number of mobile transmitters to
three while keeping the same mobility pattern as discussed above.
We can first notice that the PRR level stays high even with
increasing number of mobile nodes. This suggests that MoMoRo
adjusts itself autonomously even if the frequency of link
disconnections increases with the number of mobile nodes.
Furthermore, by observing the number of issued overhead packets,
when comparing with Fig. 10, we can notice that there is no
significant increase in overhead. This can be explained using the
following. Since mobile nodes follow the same path (e.g., a
hallway), as one mobile node
Fig. 12. Illustration of our outdoor mobile network experiment environment.
We place the data collection node at one end of a 40-m path with three mobile
nodes walking on the path while transmitting periodic packets to the collector
node with no stationary infrastructure.
TABLE II
PRR AND PACKET OVERHEAD PER NODE PER MINUTE FOR A NETWORK
WITH THREE MOBILE NODES AND NO STATIONARY INFRASTRUCTURE
experiences a link disconnection, each intermediate node to the
final destination will fix its routes to find a “currently” optimal
path. When a different mobile node experiences a disconnection at
a similar geographical region, the intermediate nodes can use the
prelearned information, thus minimizing the number of overhead
packets to perform local route recovery. Overall, our results
suggest that the performance of MoMoRo can scale with
increasing number of mobile nodes.
D. Fully Mobile Network
In this experiment, we consider three mobile nodes and one
stationary node. The three mobile nodes were carried by
humans moving at typical walking speeds back and forth on a
40-m path, as shown in Fig. 12. The three nodes were evenly
spaced out across the entire path, with at least one node in range
of the stationary data collection node at any given time. With
this experiment, we quantify the performance of a fully mobile
network in which nodes experience both changing parents and
changing children. We set the interpacket interval to 1 s and
decreased the radio transmit power to −15 dBm to achieve a
multihop network within the designated area.
We traced 5 min in total for each configuration and present the
PRR and amount of overhead packets for the three contin-uously
mobile nodes in Table II. We notice from the PRR that both AODV
and MoMoRo outperform RPL and that the over-head of MoMoRo
and AODV is higher, consistent with our pre-vious results. Not
surprisingly, the amount of packet overhead is significantly higher
given the drastic and frequent change in network topology;
however, all are well below the maximum
number of packet transmissions the radio can perform,
indicating that the nodes were able to find a parent, but not
always a parent with a proper route to the collection node. The
PRR on the other hand is significantly lower for both AODV
and MoMoRo when compared with previous results. However,
note that a new next hop selection process is triggered after a
loss of a data packet. Given that we designed the experiment to
force nodes to change parents repeatedly, the experiment was
bound to have a lower PRR than the more simpler tests.
Nevertheless, the results suggest that MoMoRo can help RPL
sustain reasonably high PRR with a relatively small increase in
overhead even in such extreme cases.
V. CONCLUSION
In this paper, we have introduced MoMoRo, a mobility
support layer for low-power wireless networks. Combined with
a data collection protocol designed for WSNs, such as RPL,
MoMoRo can effectively detect route disconnections, notify
mobile nodes of route disconnections, and robustly reconfigure
the routes. Our evaluations in an indoor testbed environment
and an outdoor setting reveal that MoMoRo can effectively
achieve high PRR performance, and when compared with other
mobile routing protocols designed for low-power wireless sys-
tems, it generates less than half the overhead packets while
maintaining similar PRR performance.
REFERENCES
[1] A. Papadopoulos, A. Navarra, J. A. McCann, and C. M. Pinotti, “Vibe: An
energy efficient routing protocol for dense and mobile sensor net-works,” J.
Netw. Comput. Appl., vol. 35, no. 4, pp. 1177–1190, Jul. 2012.
[2] M. Abolhasan, T. Wysocki, and E. Dutkiewicz, “A review of routing
protocols for mobile ad hoc networks,” Ad Hoc Netw., vol. 2, no. 1, pp.
1– 22, Jan. 2004.
[3] S. A. B. Awwad, C. Ng, N. Noordin, and M. F. Rasid, “Cluster based
routing protocol for mobile nodes in wireless sensor network,” Wireless
Pers. Commun., vol. 61, no. 2, pp. 251–281, Nov. 2011.
[4] S. Berman, V. Kumar, and R. Nagpal, “Design of control policies for
spatially inhomogeneous robot swarms with application to commercial
pollination,” in Proc. IEEE ICRA, May 2011, pp. 378–385.
[5] D. Bhadauria, V. Isler, A. Studenski, and P. Tokekar, “A robotic sensor
network for monitoring carp in Minnesota Lakes,” in Proc. IEEE ICRA,
May 2010, pp. 3837–3842.
[6] B. Chen, K. K. Muniswamy-Reddy, and M. Welsh, “Ad-hoc multicast
routing on resource-limited sensor nodes,” in Proc. 2nd Int. Workshop
Multi-Hop Ad Hoc Netw., Theory Reality, 2006, pp. 87–94.
[7] O. Chipara, C. Lu, T. C. Bailey, and G.-C. Roman, “Reliable clini-cal
monitoring using wireless sensor networks: Experiences in a step-down
hospital unit,” in Proc. 8th ACM Conf. Embedded Netw. SenSys, New
York, NY, USA, 2010, pp. 155–168.
[8] B. Ciubotaru and G.-M. Muntean, “Sasha: A quality-oriented handover
algorithm for multimedia content delivery to mobile users,” IEEE Trans.
Broadcast., vol. 55, no. 2, pp. 437–450, Jun. 2009.
[9] S. Dawson-Haggerty, A. Tavakoli, and D. Culler, “Hydro: A hybrid rout-
ing protocol for low-power and lossy networks,” in Proc 1st IEEE Inter-
national Conf. SmartGridComm, Oct. 2010, pp. 268–273.
[10] R. Fonseca, O. Gnawali, K. Jamieson, and P. Levis, “Four-bit wireless
link estimation,” in Proc. 6th Workshop HotNets, Nov. 2007. [Online].
Available: http://conferences.sigcomm.org/hotnets/2007/program.html
[11] O. Gnawali, R. Fonseca, K. Jamieson, D. Moss, and P. Levis, “Collec-
tion tree protocol,” in Proc. 7th ACM Conf. Embedded Netw. SenSys, Nov.
2009, pp. 1–14.
[12] O. Gnawali and P. Levis, The Minimum Rank Objective Function With
Hysteresis 2010, Internet Draft (Work in Progress), IETF.
[13] C. Gomez, P. Salvatella, O. Alonso, and J. Paradells, “Adapting AODV
for IEEE 802.15.4 mesh sensor networks: Theoretical discussion and
performance evaluation in a real environment,” in Proc. IEEE WoWMoM,
2006, p. 170.
[14] E. Gustafsson and A. Jonsson, “Always best connected,” IEEE Wireless
Commun., vol. 10, no. 1, pp. 49–55, Feb. 2003.
[15] D. Johnson, Y. Hu, and D. Maltz, The Dynamic Source Routing Protocol
(DSR) for Mobile Ad Hoc Networks for IPv4, Feb. 2007, RFC 4728.
[16] J. Ko, J. Lim, Y. Chen, M.-E. Razvan, A. Terzis, G. Masson, T. Gao, W.
Destler, L. Selavo, and R. Dutton, “MEDiSN: Medical emergency
detection in sensor networks,” ACM Trans. Embedded Comput. Syst.
(TECS), vol. 10, no. 1, pp. 1–29, Aug. 2010, Special Issue on Wireless
Health Systems.
[17] J. Ko, C. Lu, M. B. Srivastava, J. A. Stankovic, A. Terzis, and M. Welsh,
“Wireless sensor networks for healthcare,” Proc. IEEE, vol. 98, no. 11,
pp. 1947–1960, Nov. 2010.
[18] J. Ko and A. Terzis, “Power control for mobile sensor networks: An
experimental approach,” in Proc. IEEE Commun. Soc. Conf. SECON, Jun.
2010, pp. 1–9.
[19] J.-H. Lee, H.-W. Kim, Y.-H. Choi, Y.-U. Chung, and S.-H. Rhee, “Ip
mobility performance enhancement using link-layer prediction,” in Proc.
Future Gener. Inf. Technol., 2010, vol. 6485, pp. 171–179.
[20] S. H. Lee, S. Lee, H. Song, and H. S. Lee, “Wireless sensor network
design for tactical military applications: Remote large-scale
environments,” in Proc. IEEE MILCOM, Oct. 2009, pp. 1–7.
[21] X. Li, J. Yang, A. Nayak, and I. Stojmenovic, “Localized geographic
routing to a mobile sink with guaranteed delivery in sensor networks,”
IEEE J. Sel. Areas Commun., vol. 30, no. 9, pp. 1719–1729, Oct. 2012.
[22] D. Malan, T. Fulford-Jones, M. Welsh, and S. Moulton, “CodeBlue: An
ad hoc sensor network infrastructure for emergency medical care,” in
Proc. MobiSys WAMES, Jun. 2004, pp. 1–4.
[23] Tmote Sky, MoteIV Corporation, San Francisco, CA, USA, 2006.
[24] C. Perkins, E. Belding-Royer, and S. Das, Ad hoc On-Demand Distance
Vector (AODV) Routing, Jul. 2003, RFC 3561.
[25] C. E. Perkins and P. Bhagwat, “Highly dynamic Destination-Sequenced
Distance-Vector routing (DSDV) for mobile computers,” in Proc. Conf.
Commun. Arch., Protocols Appl. SIGCOMM, New York, NY, USA,
1994, pp. 234–244.
[26] S. K. Ray, K. Pawlikowski, and H. Sirisena, “Handover in mobile
WiMAX networks: The state of art and research issues,” IEEE Commun.
Surveys Tuts., vol. 12, no. 3, pp. 376–399, Quarter 2010.
[27] R. S. Carbajo, M. Huggard, and C. McGoldrick, “An end-to-end routing
protocol for peer-to-peer communication in wireless sensor networks,” in
Proc. 6th Workshop MiNEMA, 2008, pp. 5–9.
[28] I. Stojmenovic, Ed., Handbook of Wireless Networks and Mobile
Comput-ing. Hoboken, NJ, USA: Wiley, 2003.
[29] CC2420: 2.4 GHz IEEE 802.15.4/ZigBee-ready RF Transceiver, Texas
Instruments, Dallas, TX, USA, 2006.
[30] T. Winter, Ed., “RPL: IPv6 routing protocol for low-power and lossy
networks,” RFC 6550, Mar. 2012.
JeongGil Ko received the B.Eng. degree in com-
puter science and engineering from Korea Univer-
sity, Seoul, Korea, in 2007, and the M.S.E. and Ph.D.
degrees in computer science from Johns Hopkins
University, Baltimore, MD, USA, in 2009 and 2012,
respectively.
He was an Undergraduate Research Student under Dr.
Hyogon Kim with the Wireless Data Commu-nications
Laboratory, Korea University. He was a member of the
Hopkins InterNetworking Research Group led by Dr.
Andreas Terzis with Johns Hopkins
University. Since June 2012, he has been with Electronics and Telecommu-
nications Research Institute, Daejeon, Korea. His research interests are in
developing wireless sensing systems with ambient intelligence for the Internet
of Things and cyberphysical systems.
Marcus Chang received the Ph.D. degree in com-
puter science from the University of Copenhagen,
Copenhagen, Denmark.
He is currently a Postdoctoral Fellow with Johns
Hopkins University, Baltimore, MD, USA, where he
teaches embedded systems and builds wireless
sensor networks for environmental monitoring. His
research interests are in distributed computing, wire-
less networks, and cyberphysical systems.