The Wireless Autonomous Spanning tree Protocol for Multihop Wireless Body Area Networks

Abstract

Wireless body area networks (WBANs) have gained a lot of interest in the world of medical monitoring. Current implementations generally use a large single hop network to connect all sensors to a personal server. However recent research pointed out that multihop networks are more energy-efficient and even necessary when applied near the human body with inherent severe propagation loss. In this paper we present a slotted multihop approach to medium access control and routing in wireless body area networks, the wireless autonomous spanning tree protocol or WASP. It uses crosslayer techniques to achieve efficient distributed coordination of the separated wireless links. Traffic in the network is controlled by setting up a spanning tree and by broadcasting scheme messages over it that are used both by the parent and the children of each node in the tree. We analyze the performance of WASP and show the simulation results

Full text

Page 1

The Wireless Autonomous Spanning tree Protocol
for Multihop Wireless Body Area Networks
Bart Braem1, Benoˆ ıt Latr´ e2, Ingrid Moerman2, Chris Blondia1, Piet Demeester2
1University of Antwerp – IBBT vzw – Department of Mathematics and Computer Science – PATS
Middelheimlaan 1, B-2020 Antwerpen, Belgium, {bart.braem, chris.blondia}@ua.ac.be
2Ghent University – IBBT vzw – IMEC vzw – Department of Information Technology (INTEC) – IBCN
Gaston Crommenlaan 8, B-9050 Ghent, Belgium, {benoit.latre, ingrid.moerman, piet.demeester}@intec.ugent.be
Abstract—Wireless body area networks (WBANs) have gained
a lot of interest in the world of medical monitoring. Current
implementations generally use a large single hop network to
connect all sensors to a personal server. However recent research
pointed out that multihop networks are more energy-efficient and
even necessary when applied near the human body with inherent
severe propagation loss. In this paper we present a slotted
multihop approach to medium access control and routing in
wireless body area networks, the Wireless Autonomous Spanning
tree Protocol or WASP. It uses crosslayer techniques to achieve
efficient distributed coordination of the separated wireless links.
Traffic in the network is controlled by setting up a spanning
tree and by broadcasting scheme messages over it that are used
both by the parent and the children of each node in the tree.
We analyze the performance of WASP and show the simulation
results.
I. INTRODUCTION
In the last few years, wearable health monitoring systems
have gained the attention of various researchers in order to
cope with the rising costs of the health care system. An
important task of such a system is to collect physiologi-
cal parameters like the heartbeat, body temperature, SpO2-
levels,...To this purpose, several sensors are placed in clothes,
directly on the body or under the skin of a person. If these
sensors are equipped with a wireless interface, a new type of
network spanning the entire body takes form: a wireless on-
body network or a Wireless Body Area Network (WBAN) [1].
A WBAN provides continuous health monitoring and real-time
feedback to the user and the medical personnel. Furthermore,
the measurements can be recorded over a longer period of
time which improves the quality of the measured data [2].
Next to sensor devices, actuators or actors are used. These take
some specific actions according to the data received from the
sensors or through interaction with the user. For example, an
actuator can be fitted out with a built-in reservoir and pump for
administering the correct dose of insulin to diabetics based on
the measurements of glucose level received from the sensors.
Next to these sensors and actors, each WBAN has a sink or
personal server such as a PDA. It is attached to the hip or
wrist and is responsible for the interaction with the user.
Current research efforts on WBAN largely focus on the
development of new sensors and new radio interfaces. One
of the major concerns is to lower the energy usage of the
different devices in order to extend the life time and therefore
the user-friendliness of the system. Several methods exist to
achieve this purpose: the development of low-power radio
interfaces and low-power sensors and the use of optimized
network protocols which lower the energy consumption even
further. Apart from medium access control another major point
of focus within WBAN research are propagation issues. It was
found that radio signals experience severe path loss close to
the body [3]. Consequently, direct communication between all
nodes will come at high transmission cost or will prove to be
impossible.
In this paper, we introduce a new approach for communi-
cation in multihop wireless body area networks: the Wireless
Autonomous Spanning tree Protocol or WASP. This protocol
incorporates slotted medium access and the automatic setup
of the different routes. To that end, a spanning tree is set up
automatically. This spanning tree is subsequently used to route
the data toward the sink and to assign the different slots in a
distributed manner. Thus, WASP uses a crosslayer approach
where medium access control and routing are handled by the
same spanning tree and therefore a higher throughput and
lower energy consumption can be achieved.
The remainder of this paper is organized as follows. An
overview of related work is given in section II. In section III
we will give a general overview of WASP by using a small
example network. An analysis of the performance is given in
section IV. In section V the implementation and experimental
results of the protocol are stated. Finally, section VI gives
some future directions and section VII concludes the paper.
II. RELATED WORK
Protocols for WBANs can be divided in intra-body commu-
nication and extra-body communication ones. The first control
the information handling between the sensors or actuators
and the sink, the latter ensure communication between the
sink and an external network. Doing so, the medical data
of the patient at home can be remotely consulted by a
doctor. The WASP-protocol addresses the issue of intra-body
communication, therefore we will focus this section to this
type of communication.
In the last year, several implementations of WBANs have
been proposed. In [4] a system architecture of a wireless body

Page 2

area sensor network is presented. This system architecture
both handles the intra- and extra-body communication. The
communication between the sensors and the sink is slotted.
The slots are synchronized using beacons periodically sent
by the sink. A similar system is proposed in [1] where the
authors use IEEE 802.15.4/ZigBee compliant radios. This type
of radio is also used in the CodeBlue-project [5] and in [6].
The BASUMA-project (Body Area System for Ubiquitous
Multimedia Applications) [7] aims at developing a full plat-
form for WBANs. As communication technique, a UWB-
frontend is used and a MAC-protocol based on IEEE 802.15.3.
This protocol also uses time frames divided into contention
free periods (with time slots) and contention access periods
(CSMA/CA). In [8] a collision free real-time MAC-protocol is
implemented. This protocol divides time into frames in which
only one node is allowed to transmit. The scheduling order is
derived by applying the Earliest Deadline First algorithm.
Most of the projects above only consider direct communi-
cation between the sensors and the sink. However, the prop-
agation loss around the human body is very high. Generally,
in wireless networks, it is known that the transmitted power
drops off with dnwhere d represents the distance between
the sender and the receiver and n the path loss coefficient.
In free space, n has a value of 2. Several researchers have
been investigating the path loss around the human body [3],
[9]–[12]. All of them come to the conclusion that the radio
signals experience large losses as n reaches a value between 4
and 7. Hence, direct communication between the sensors and
sink will not always be possible, especially when one wants
to lower the transmission power of the radio in order to save
the precious energy available in the nodes. The projects above
do not consider multihop communication.
A lot of research has been done in the area of sensor
networks. As body area networks share some characteristics
with this type of network, we will consider some techniques
where multihop communication is used. Radio energy con-
sumption in sensor networks is reduced by controlling the
power and duty cycle of the radio. Scheduled protocols,
such as S-MAC [13], uses scheduling to coordinate sleeping
among neighboring nodes to avoid idle listening. A preamble
sampling technique is used in WiseMAC [14] where a node
regularly samples or polls the medium for a very brief time
to check whether a packet needs to be received. An ultra-low
duty cycle MAC that combines scheduling and channel polling
is presented in [15]. Another technique is TDMA where
contention-introduced overhead and collisions are avoided.
III. PROTOCOL DESCRIPTION
A. General overview
Sensors in a WBAN can be considered as generating CBR
traffic where traffic requirements do not change quickly over
time [16]. Therefore WASP uses coordination of the medium
access.
WASP is a slotted protocol that uses a spanning tree for
medium access coordination and traffic routing. Each node
will tell its children in which slot they can send their data
by using a special message: a WASP-scheme. This WASP-
scheme is unique for every node and constructed in the node
sending the scheme. A distinguishing property of WASP is the
dual usage of the WASP-schemes by exploiting the broadcast
nature of wireless links. A node uses the schemes to control
the traffic of its children and simultaneously to request more
resources from its parent for these children. This minimizes
the coordination overhead because each scheme is used by the
parent and the children of the sending node. Everything the
node has to know to generate this scheme can be obtained by
listening to the WASP-schemes coming from its parent node
(i.e. one level up in the tree) and from its children (i.e. one
level lower in the tree). Consequently, the division of the time
slots is done in a distributed manner.
Hereafter a short overview is given of the operation of the
protocol in a steady-state situation. In this section we assume
a spanning tree has already been constructed and is used
to propagate the data to the sink. The time-axis is divided
in different cycles (further referred to as WASP-cycles) that
consist of a fixed number of time slots. In a WASP-cycle, each
node is allowed to send its data and/or to forward data received
in the previous cycle to the next node. At the beginning of
each cycle, the sink sends its WASP-scheme to its children.
This WASP-scheme informs the sink’s children when they can
send their information. These children respond to the scheme
by sending out their own WASP-scheme in their designated
time slots. These WASP-schemes are based on the WASP-
scheme of the sink and on the requirements of their children.
Thus each node right below the sink calculates its own WASP-
scheme and sends it to its own children which form the second
level. On their turn, these nodes send out the WASP-scheme
and so on. Doing so, each node will know when it can send
its data without the need for a device that centrally calculates
the distribution of the time slots. A more elaborate example
will be given in section V-B.
B. WASP-schemes
Each node will send a WASP-scheme to its children to
inform them when they are allowed to use the link. In this
section, we will describe what elements can be found in such
a WASP-scheme. In order to focus our thoughts, the simple
example network of figure 1 will be used. The lines between
the nodes indicate the tree structure of the network. All
communication is wireless and the tree is constructed in such
a way that the nodes only can hear their parent, their siblings
(i.e. children of the same parent) and their children. This
assumption simplifies the protocol explanation. We assume
that in this example each node only has one data packet to
send to the sink per cycle. Of course, in general, nodes will
be allowed to send multiple data packets per cycle.
The WASP-scheme of the sink (node S) is as follows:
1
????
S AB
????
2
3
????
.3ABB
? ?? ?
4
5
????
X 11101
? ?? ?
6
(1)

Page 3

Slot
0
1
2
3
Level 0
S: WASP-scheme: SAB.3ABBX-10011
Level 1Level 2
A: send data + WASP-scheme: A.1CX-1
B: send data + WASP-scheme: BDEX-11
C: send data + WASP-scheme: C.1X-
D: send data + WASP-scheme: D.2X-
E: send data + WASP-scheme: E.1X-
Contention slot C
Contention slot D
Contention slot E
4
5
6
Contention slot A
Contention slot B
A: forward data
7
8
9
B: forward data
B: forward data
Contention slot S
Fig. 2.Steady-state WASP cycle for the example network
S
A
B
C
Level 0
Level 1
Level 2
E
D
Fig. 1. Example network. The lines indicate the tree structure.
where
1
2
=
=
The address of the sending node;
Every child of the sink is designated one slot.
In that slot, the node sends its WASP-scheme
followed by data;
A silent period of 3 slots. In this period, the child
nodes receive data from their children that they
need to forward to the sink;
The children forward the received data to the
sink;
Contention slot;
ACK-sequence.
First, each of the child nodes of the sink is awarded 1
slot, part 2 of equation (1). In this slot, the child nodes
will send their WASP scheme and data. This WASP-scheme
will be heard by the sink and by the children of the node.
Consequently, the sink is informed of its child node’s traffic
requirements and is capable of calculating the duration of the
silent period for the next cycle.
During the silent period, part 3 of equation (1), the sink
will not be sending or receiving any messages as this period
is used by the sink’s children to receive data or messages from
their children. Doing so, interference between different levels
will be limited.
Following the silent period, a sequence is given in which
the children can forward the received data to the sink, part
4 in (1). After that, a contention slot is inserted in order
to allow new nodes to join the network, part 5 in (1). The
scheme is completed by the piggybacked acknowledgments,
represented by simple bits, part 6 in (1). This will be explained
3=
4=
5
6
=
=
in section III-H.
A slightly different WASP-scheme is used by the other
nodes, for example node A:
1
????
A
3
????
.1
C
????
4
5
????
X
1
????
6
(2)
A first difference is the duration of the silent period. Whereas
with the sink the silent period was used to allow the nodes of
level 2 to send their data to level 1, the silent period is now
used to indicate the nodes of level 2 when they should remain
silent in order not interfere with the ongoing communication
on the higher level. In other words, in the silent period of the
sink, the sink will be silent, and in the silent period of the
other nodes, their child nodes will be silent. This reasoning
brings us to the second difference with (1): the omission of
part 2. The nodes below level 1 are not allowed to send their
data immediately as all siblings of their parent node need to
send their WASP-scheme first in order to reduce delay.
In the other levels (level three and lower), the nodes can start
sending when the transmission of the parent node has ended,
but the same principle holds: the silent period indicates when
the child nodes should remain silent (and consequently can
turn off their radio). A more elaborate example with a larger
number of levels can be found in section V-B.
An overview of the entire WASP-cycle of this example can
be found in figure 2. In each timeslot, it is shown what action is
taken. In timeslot 0, the sink S, at level zero, sends its WASP-
scheme. In the next timeslot, sensor A sends its WASP-scheme
and its data. Further, in timeslot 3, we can see that node C
and D can send simultaneously.
C. Duration of the silent period
The duration of the silent period for the sink SPS equals
the number of slots needed to send the data from level two to
level one. Thus, it can be calculated as follows:

SPS=
max
∀i∈ChS
?
j ∈Vi
TSj

+ 1
(3)
where Chiare the children of node i, TSj is the number of
slots needed by node j to send its own data and Videnotes the

Page 4

nodes in the tree below node i, node i not included. One extra
slot is inserted in order to allow the presence of a contention
slot.
The length of the silent period of the other nodes is
calculated based on the parent’s WASP-scheme and thus is
different below level 1:
SPlevel1
=
slot # start silent period -
slot # node’s first occurrence - 1 (4)
slot # contention slot -
slot # node’s first occurrence
SPleveli > 1
=
(5)
For example, the length of the silent period of node A equals
the slot number of the start of the silent period (= 3) minus the
slot number of the first occurrence of node A in the WASP-
scheme of the sink (= 1) which gives a silent period of 1 slot.
The length of the silent period of node C is the slot number
of the contention slot (= 4) minus the slot number of the first
occurrence of node C (= 3) which equals 1 slot. Indeed, a
silent period of one slot is necessary because node A still has
a contention slot.
D. Contention slots
A contention slot allows nodes to join the network by
sending a JOIN-REQUEST. It is possible to omit this slot
for a few consecutive WASP-cycles to increase the maxi-
mum throughput. However, in order to keep the connection
time reasonably low, at least one contention period per n
milliseconds is required. Moreover, to support a notion of
mobility a high frequency of contention slots is required. As
it is possible that more than one node can send during these
contention slots, a random delay is inserted before each node’s
transmission. Doing so, the probability of collisions during
these slots is decreased. There is no special mechanism that
detects collisions. If a node that wants to join the network does
not hear its address in the next WASP-scheme of its parent, it
assumes that a collision has occurred and sends a new JOIN-
REQUEST in the next contention slot.
E. Heterogeneous data rates
In the example above, each node was only assigned one data
packet or slot per cycle. In more general networks, some nodes
will require a larger data rate then others. In such cases, these
nodes will be assigned multiple slots per cycle. The desired
number of slots can be requested when joining the network.
Sometimes the wanted data rate of a node can change over
time, e.g. when more accurate measurements are desired over
a certain period of time. The node will ask for more slots in
its WASP-scheme. This will be seen by the parent node that
will assign more slots in its next WASP-scheme.
F. Building and maintaining the spanning tree
The following steps are taken to let nodes join the network:
1) The new node scans the wireless medium for a certain
time and picks up the WASP-scheme of the nodes in
range.
2) Based on these received WASP-schemes, the new node
determines which node will be the best parent, based
upon the node’s requirements (low delay, reliability, a
weighted average, ...).
3) During the contention slot of the chosen parent, the new
node sends a JOIN-REQUEST.
4) The parent receives this request and adds the new node
in the next WASP-scheme.
If for a certain time a child node does not receive a WASP-
scheme, a new parent is chosen using the same procedure.
Each node should periodically check which nodes are in its
neighborhood and choose a new parent if a better one is
around. The node does not have to notify its parent. If a parent
doesn’t receive the WASP-scheme of a child for n consecutive
times, it assumes that the child is no longer part of the network
and doesn’t include it in the WASP-scheme anymore.
These basic tree maintenance mechanisms support low
mobility.
G. Routing and addressing
In order to facilitate the forwarding to the different nodes,
the tree structure of the network can be used for addressing.
However this poses problems as nodes might change attach-
ment points in the tree. A more general solution is manually
asking the sink for a unique address. The sink receives traffic
from all sensors so it knows the addresses in use. A WASP
address is composed of 6 bits. The sink gets address 000000,
the first node will get 000001 and so on. Using 6 bits the
number of nodes is limited to 64, which is reasonable due to
the nature of a BAN (see section IV-D).
In a WASP scheme 1 byte describes each slot. The first bit
denotes the slot type. 0 stands for a regular slot where data is
sent and 1 for a special slot. In case of a data slot the second
bit denotes the traffic direction: 0 for regular traffic to the sink
and 1 for traffic in the other direction or more general traffic
that requires routing. The other 6 bits define the actual address
of the node that is permitted to send in the slot. In case of a
special slot, the 7 remaining bits are used to define the length
of the silent period. E.g. 1.0000011 denotes a silent period of
length 3. If all the bits are set to 1, the slot is a contention
slot instead of a silent period.
Most traffic in a WBAN flows from the nodes to the sink.
Traffic in the other direction can be supported by setting the
second bit in the WASP-scheme. A node will add the source
address with that bit set to the WASP scheme. Each child then
decides whether it is on the path to the destination. If so it
turns on its radio to receive the packet. Routing can be done
by using a technique similar to learning bridges. Nodes record
the addresses in traffic passing by and route packets from the
sink to the nodes using that information.
H. Acknowledgments
At the end of each WASP-scheme, an ACK-sequence for
the previous WASP-cycle is sent. It contains a bit for each
slot in which data was sent in the previous cycle. A 0 denotes
that the packet was not received correctly, a 1 denotes success.

Page 5

The node that receives the WASP-scheme, say node N, will
check the ACK-sequence. If the position of a 0 corresponds
to one of the slots where N sent data, the node will resend
that data in this cycle. Its parent will already include an extra
slot for node N.
This acknowledgment scheme is quite weak but it suffices
largely. Due to the absence of contention data loss will be
limited to interference problems.
I. Synchronization
The nodes in the tree need to be synchronized in order to
avoid shifting of the start of slots between nodes. However the
recurring cycles allow for resynchronization at the beginning
of each cycle. Each node should wake up some milliseconds
before the start of slot to avoid these issues.
IV. PERFORMANCE ANALYSIS
In this section, we present the performance analysis of our
proposed protocol. We will address performance issues such
as minimum delay, maximum throughput and sleeping time of
the nodes.
A. Maximum overall throughput
The maximum overall throughput of the protocol can be
seen as the percentage of the useful traffic (or goodput) that
can be sent over the network. As we are working in a multi hop
environment, the maximum throughput will mostly depend on
the number of nodes in the network and the number of levels
in the tree. Indeed, the maximum amount of data that can be
sent to the sink per WASP-cycle depends on the number of
nodes in the first level and the length of the silent period.
As mentioned in section III-B, the silent period of the sink
allows the nodes of level one to receive their data. This
means that if we can lower the duration of the silent period,
the maximum throughput will rise. Thus by minimizing the
number of children of the nodes of level one, the maximum
throughput can be improved. Generally, the throughput in
terms of percentage can be found as
# slots where data sent to sink
length of a WASP-cycle
where the length of the WASP-cycle is expressed in slots and
depends on the length of the silent period.
In the example of figure 1, we see that per WASP-cycle
5 packets can be sent to the sink and the total length is 10
timeslots. Thus, we have a maximum throughput of
This seems to be a low number, but we have to keep in mind
that we are working in a multi hop environment.
The following formula determines the length of a WASP-
cycle TWC:
TP =
(6)
5
10or 50%.
TWC
=
# data children L1+ SPS +
forwarding data of L2from L1to L0+ 2
(7)
where Li represents level i. The two extra time slots added
at the end are used for the transmissions of the sink’s WASP-
scheme and the contention slot at the end.
The duration of the forwarding period equals the number of
timeslots needed to send the data of each node. Using (3), we
can rewrite this formula as
TWC =
?
i∈V
TSi+ max
∀i∈ChS

?
j ∈Vi
TSj

+ 3
(8)
Thus, (6) can be reformulated as
TP =
?
i∈VTSi
?
i∈VTSi+ max∀i∈ChS
??
j ∈ViTSj
?
+ 3
(9)
and highly depends on the number of nodes and the struc-
ture of the tree. In order to evaluate this formula, we will
distinguish two extreme cases: 1) all nodes can communicate
directly with the sink and 2) all nodes are in different levels.
We further assume that each node only has 1 packet to send per
WASP-cycle. This means that in both cases, the nominator of
(9) equals the number of nodes in the network (x). In the first
case, the second term in the denominator of (9) equals zero
and in the second case to x−1. The boundaries for 50 nodes
are 94% in the first case and 49% in the second case. The
latter has the largest silent period. Thus, in order to increase
the throughput, the silent period should not be too long. This
can be achieved by setting up a balanced tree.
B. Delay limits
The experienced delay depends on the number of levels
present in the network. Indeed, a node can send his data only
up one level during each WASP-cycle. The only exception is to
be found at level 1, where the sink’s children can first receive
the data from their children and then forward the data.
We can define an upper and lower bound for a node i:
Lower bound = max((level nodei− 2) · TWC,1)
Upper bound = max((level nodei− 1) · TWC,1)
The maximum function is needed as the delay can not be lower
than 1 slot.
The maximum delay over the whole network can be ex-
pressed as follows, assuming that the network has at least 2
levels:
?
Summarizing, if we want to have a high throughput, we
should minimize the length of the silent period and for a low
delay minimize the number of levels. These two conditions
do not contradict, therefore a high throughput can be achieved
while preserving the low delay.
(10)
(11)
maximum delay =max
∀i∈V(level nodei) − 1
?
· TWC. (12)
C. Sleeping nodes
When the nodes have heard the WASP-scheme of their
parent, they can turn their radio off in the slots where they
are not involved in the communication. This allows for energy
saving. In the example scheme of figure 2, the sink can turn
its radio off in the silent period as it will not receive data