Conference PaperPDF Available

Energy balanced routing for latency minimized wireless sensor networks

Figures

Content may be subject to copyright.
Energy Balanced Routing for Latency Minimized
Wireless Sensor Networks
Hans-Peter Bernhard, Andreas Springer
Johannes Kepler University Linz
Institute for Communications Engineering
and RF-Systems
Altenberger Str. 69, 4040 Linz, Austria
Email: h.p.bernhard@ieee.org
Peter Priller
ITS Global Research & Technology
AVL List GmbH
H. List Platz 1
8020 Graz, Austria
Email: peter.priller@avl.com
Leander B. H¨
ormann
Sensors & Communication
Linz Center of Mechatronics GmbH
Altenberger Str. 69
4040 Linz, Austria
Email: leander.hoermann@lcm.at
Abstract—We present a communication and routing pro-
tocol to balance energy consumption among wireless sensor
and actuator nodes under the constraint of minimizing end-
to-end latency. Minimized and nearly deterministic end-to-end
latency allows centralized real time data recording and actuator
control. Moreover, defined latency improves security and safety
mechanisms to come closer to the trustworthiness of a cable while
realizing the advantages of wireless networks. Battery powered
or energy harvesting wireless nodes are inevitable to avoid power
lines and therefore, we cope with limited energy budgets. In order
to balance the communication load among sufficiently supplied
nodes, the energy budget is part of the routing path metric. We
show that the time domain multiple access scheme is crucial to
enable defined latency and improve energy efficiency by avoiding
collisions. If the energy distribution among the nodes is known,
this novel protocol allows to calculate latency and round-trip time
in advance. An asymmetric timing scheme can limit multi-hop
latency for either upstream or downstream communication by
one superframes (SF) duration. The presented ruleset is able to
reconfigure the routing focused on energy constraints of involved
nodes.
I. INT RODUCT IO N AN D RE LATED WORK
In this paper we present a communication and routing
protocol for low power sensor networks in limited areas,
e.g. 100 m x 100 m, using Bluetooth®low energy (BLE) as
physical (PHY) layer. In many low and ultra low power
sensor network technologies such as BLE, routing concepts
are missing but necessary, if network coverage needs to be
extended. In industrial wireless sensor and actuator networks
(IWSANs) robustly synchronized low latency networks are
demanded in addition. IWSANs should be able to stream
sensor data from autarkical sensors in the same way as from
sensors being attached to a measurement unit via cable [1].
BLE can serve as PHY layer due to its low power consumption,
but the communication range of one base station is limited by
the low transmit power of the nodes. To increase the coverage,
routing over a few hops can be introduced, which is very
effective because the coverage is increased proportionally to
the squared number of hop distances.
IWSANs have requirements that are much more demanding
than general purpose wireless sensor networks. The most
prominent ones are dependability (sensor data must not be
lost), synchronicity (sensor data need to be sampled and
transmitted synchronously on all nodes), deterministic and low
latency, and a very high security level. In most applications,
these requirements need to be fulfilled under the constraint of
a severely limited energy budget. If multi-hop communication
is introduced in an IWSAN, fulfilling the aforementioned
requirements becomes even more demanding and is only
possible if the routing protocol is designed accordingly.
Routing is a problem to which many solutions exist in
literature. The enormous number of different algorithms [2]
reflects that routing is heavily depending on the application
and the specific environment. In our work routing is solved
under two main constraints: energy efficiency and latency. If
energy efficiency is considered with routing, load balancing
is one major concept to achieve this [2]. Standard routing
algorithms find all routes in the network by searching route
by route between two different nodes. Here, we present a
different way of solving the routing problem while minimizing
energy consumption and latency. We assume a static IWSAN
with a single base station (BS) and synchronized nodes which
communicate using a time domain multiple access (TDMA)
based medium access control (MAC) protocol. The network
is centrally managed, which is a reasonable assumption for
industrial applications. First of all, layers are defined and all
nodes are assigned to a layer. All routes start at the BS.
The optimized routing task is subdivided into layer-by-layer
routing similar to [3]. In each layer routing nodes are found
by distributing the routes according to the routing capability of
each node, which also includes its energy budget. The layers
are processed sequentially beginning with the node or layer
furthest away from the BS. In [2] a wide variety of protocols
are mentioned and we use the results to select a fruitful
combination of hierarchical and other optimization tweaks.
Other published load balancing methods rely on ad hoc
on-demand multipath distance vector (AOMDV) [4], which
are not optimized for a static network and which use non-
deterministic packet queuing. This does not fulfill the require-
ment of deterministic latency. In [5] the packet header is ex-
tended by additional energy budget information and the routing
is packet-based and not centrally managed. Other approaches
like [6], [7] are using some heuristic function to balance
the routing among the repeating nodes. In [8] the Dijkstra
algorithm is modified to avoid routing bottlenecks by adjusting
the node weights. Energy balancing is also considered in [9]
but end-to-end latency is no objective of this work.
In contrast to [6], we solve the minimum latency issue by
organizing the network in a layered structure, which organizes
the nodes in a hop hierarchy defined by the link quality. So,
the minimum hop distance is the result of a measurement
and not an optimization. The approach of low-energy adaptive
clustering hierarchy (LEACH) [10], [11] clusters the wireless
network to achieve energy optimization with randomly selected
cluster heads. This introduction of randomness is not able to
fulfill the requirement for deterministic latency. In contrast
to [12] we do not set the weight of one node in relation to
the network wide average energy. Instead, we compare the
route metric from the energy perspective only within equal hop
distances. In contrast to other routing mechanisms, a change
of routes does not change the latency in our approach.
The remainder of the paper is organised as follows. In
section II the network discovery and the routing algorithm
are presented. The communication protocol and its necessary
extension to implement the routing is explained in section III,
section IV deals with the energy model and presents data from
hardware demonstrators, and finally conclusions are drawn in
section V.
II. ROUTING
The introduced routing mechanism is set up to fulfill the
following objectives:
centrally planned routing
energy balanced routing
minimizing latency
collision free access scheme
It allows to balance the set of routes with respect to their
energy consumption instead of individually optimizing route
by route. Routing is performed by a two pass algorithm, first
discovering the network structure and second setting the routes.
To find the routes, all calculations and optimizations are done
at the central BS which is mains powered. The nodes just have
to gather network structure, link quality and energy budget
information. The BS can collect energy information during
regular communication to monitor the energy situation. If the
energy situation indicates that a re-routing is required, the BS
is able to trigger a new routing procedure. This method is
optimizing energy consumption over all existing routes. Partly
re-routing procedures are thus not possible with the same
quality. Hence, this routing mechanism works best for fixed
IWSAN networks.
A. Network structure
Let us assume, that all considered nodes Nkare known to
the BS and are elements of the set R={N0, N1, . . . , NK}.
N4
N11 N20
N10
N1
N2
N3
N21
N22
BS
N12
R1
R2
R3
Fig. 1: Layered network structure
Additionally, all nodes are managed by the BS which is
assigned to N0. The network structure reflects this in its
simplest form by using a star topology. There is always a
non-zero probability that not all nodes have a reliable direct
link to the BS because of limited transmit power or other
restrictions. Hence, nodes are used as repeaters to extend
the network, in order to allow an increased coverage. Under
the constraint of minimizing end-to-end latency the repeating
nodes themselves are used as hubs of a star sub-topology.
Consequently, the resulting network topology is a clustered
star. As a side effect, the minimization of the number of
hops avoids loops in the routing path by the layered network
structure. Based on this centralized structure, it is not necessary
to implement the spanning tree protocol (STP) [13] for wireless
or mobile applications as in [14]. Nevertheless, the problem
of selecting shortest path algorithms for load balancing has to
be solved. Similar to [15], [16] we use other qualifying path
properties in addition to the standard path metric (cf. section
IV).
To implement the communication required for the network
discovery we use the energy and power efficient synchronous
sensor network (EPhESOS) protocol [1] (cf. section III). The
EPhESOS protocol has no routing strategy and assumes a
star topology with one base station, but for our envisioned
application the network structure is not predefined. Therefore,
the routing process consists of two parts. First the network
structure of the nodes needs to be discovered. This process has
to be energy efficient. Therefore, we assume that the clocks
of the nodes are synchronized at a coarse level, e.g. minutes,
and the nodes are taking part in the discovery process at a
predefined time. Another method to wake up and take part in
the discovery process could be passive RFID listening [17],
[18], [19]. In a second step the routes are computed and
communicated to the nodes.
B. Network discovery
The network structure is built up by using an iterative
strategy which successively connects the nodes to the network.
Algorithm 1 BAckward Structured Energy BAlanced Link
(BASEBAL) Network discovery
1: function GE TST RUCTUR E(R)
2: Rf=R\N0
3: n0
4: Rn={N0}
5: while Rn6=do
6: for all NkRndo
7: for all NcRfdo
8: if Ncresponds to Nkthen
9: Rn+1 Rn+1 Nc
10: FNkFNkNc
11: BNcBNcNk
12: RfRf\Rn+1
13: nn+ 1
14: nmax n1
15: return nmax,Rp,FK,BK
The BS is transmitting periodic beacons used as synchroniza-
tion reference (cf. section III-A). As depicted in Fig.1, those
nodes acknowledging the beacon of the BS represent the set
R1R. The nodes in R1provide a minimum communication
link quality to the BS. The other nodes which are not reached
directly by the BS or have a very poor communication link
quality, are part of R\R1shown in Fig. 1 as R2R3. In a
next step, nodes R1send out beacons themselves. Nodes
which are able to respond to beacons from nodes R1with
a sufficient link quality are therefore assigned to R2. Nodes
R2need one intermediate hop to reach the central BS. The
iterative procedure can be formulated as
Rn+1 = (R\ ∪n
k=1 Rk)respond to Rn(1)
where respond to selects those nodes, contained in the set of
the left-hand side, which respond to some or all node elements
of the set Rn.
The network discovery and the computation of routing ta-
bles, forward and backward links is described into more detail
in Algorithm 1, which presents the function GE TSTRU CT UR E
used with the parameter R. As already defined, Ris the set of
all available nodes and obviously equal to all free nodes Rf
in an initialization step. Free nodes are all nodes not assigned
to one specific layer. Furthermore, the set R0is set to contain
only the BS. As long as Rnis a non-empty set, the algorithm
processes all nodes of Rn. Each node NkRnis a possible
routing node. To discover to which nodes from Rfit can
communicate, it transmits a beacon and switches into receiving
mode for at least 2 wakeup periods. A wakeup period TWAP is
the period within nodes wake up during the network discovery.
In response to the beacon from node Nk, the free nodes from
Rfare sending out a message and receive as answer time slots
and a synchronization beacon as implemented in the EPhESOS
protocol. We have shown in [1], that two wake up periods
are sufficient to detect all reachable nodes in a network of
100 nodes as reasonable size. The BS triggers the beacon
transmission of the nodes from Rnand keeps track of the
newly detected nodes by adding them to Rn+1. Additionally,
Algorithm 2 BAckward Structured Energy BAlanced Link
(BASEBAL) Network link phase
1: function SE TRO UT ES (nmax,Rp,FK,BK)
2: for n=nmax 1; n > 0; nn1do
3: for all NkRndo
4: Fu
NkFNk\ S
NgRn\Nk
FNg!
5: T← ∅
6: Lmin{max#BNk,#Nk}
7: for all Ng∈ {Fu
Nk,FNk\Fu
Nk}do
8: if L > #T+ #Ngthen
9: T← {T, Ng}
10: FNkT
11: #Nk#FNk+ 1
12: for all NgRn\Nkdo
13: FNgFNg\FNk
14: BNgBNg
\min
#NiNiBNg#Ni>#T
15: return Rp,FK
each detected node Ncis added to a forward list FNkfor
each routing node Nk. Vice versa, the routing node Nkis
added to the backward list BNcof the newly detected node.
This procedure is applied as long as Rnhas nodes that are
responding for the first time. For each layer the set of free
nodes is reduced by removing Rn+1 from Rf. As mentioned
before, the algorithm runs as long as nodes exist in a new layer
of the network. If the algorithm has stopped, the maximum
layer number nmax is found.
With (1) implemented as Algorithm 1, we split up the set of
all nodes Rinto disjunct subsets Rp={R1,R2,...,Rnmax }.
Each of these sets have a routing distance to the BS equivalent
to their subscript. The BS has to be connected to all Knodes
and therefore Kroutes have to be found. Moreover, the set of
nodes Ris partitioned in Rpwhere all layers represent one
element of the partition. FKand BKrepresent the set of all
forward and backward nodes as result of the network discovery
algorithm.
C. Network routes
All Kroutes start at BS and each route connects one node
out of R. First of all, we consider latency as weight, therefore
all the weights are positive and thus we can apply all simple
routing algorithms. It is evident, if we use e.g. Dijkstras’s
algorithm to each route, that each possible route is optimal in
the sense of latency. The Kindividual optimal routes can have
different number of hops to different nodes, but this number
is constant per route as long as each layer provides at least
one connection to reach all nodes of the next layer. The hop
numbers per route range from 0 to nmax 1and all routes have
to be maintained synchronously. Therefore, all nodes within
the network can send and receive data from the BS with a
defined latency. The dominating source of the latency is the
delay introduced by the time between receiving and resending
at the routing nodes. As the latency of an individual route is
always optimal in this network structure, the energy budget of
the routing nodes can be used for further improvements.
In section IV the energy management is discussed in
more detail. Here we simply apply the so-called link energy
equivalent (LEE), which represents the number of links that
can be processed by one node due to its energy budget. We
have to take care that this budget is not exhausted by too
many routes via one node. The energy needed for one link is
given in (6) and the LEE values can be found from Fig. 6.
The core element of the routing Algorithm 2 is the function
SE TRO UT ES , which processes the discovered network char-
acterized by nmax,Rp,FK,BK. The function S ET ROU TES
implements the two key ideas of the algorithm: I) It starts with
layer nmax and proceeds backwards to the BS. II) The link
requests of layer n+ 1 are distributed according to the energy
budget of the individual nodes in layer n. Therefore, we
name the algorithm backward structured energy balanced link
(BASEBAL) algorithm.
The algorithm starts with the set of nodes Rn=Rnmax 1
which is sorted by the LEE values of the nodes. In a simplified
version the algorithm in the loop over all nodes Nkin Rntakes
the node with the highest LEE value and considers all nodes
from Rk+1 which are linked to it. The number of linked nodes
is now limited by the LEE value of the node itself and limited
by the highest LEE value of layer k1. Then, all connected
nodes of layer k+ 1 are marked as routed. However, the
implemented loop is more complicated as we have to consider
several link restrictions. This is described in full detail in the
next paragraph.
As mentioned already, the algorithm starts with the layer
n=nmax 1and processes all nodes Nkin Rn. All routes,
which have only Nkas routing node in layer n, have to be
handeled first. As an example, consider node N31 R4in
Fig. 2. This node has only one possible route to layer R3via
node N21. Hence, this route has to be treated with priority
which means, before other nodes R4which have more than
one routing node in layer R3. Generally, Fu
Nkis the set of
all nodes in layer n+ 1 having only one possible route to the
layer Rnand this route is via node Nk. It is calculated by
taking the union of all forwarding nodes except those from
Nkas S
NgRn\Nk
FNg. In our example this union of forwarding
nodes has one element N30. The forwarding nodes of N22
are FN22 ={N30, N31 }. If the union is subtracted from this
set, N31 is left over as node, that has no alternative except
the route by N22. The evaluation of the number Lof links
routable by this node is processed next. Lis calculated as the
smallest number from either #Nk, which is the LEE value, or
max#BNk, which is the maximum LEE value of all possible
backward nodes.
The next loop takes all nodes of Fu
Nkwith priority and
then the rest of FNkand checks whether the link count #Ng
plus the already linked nodes in the temporary set Tis less
than L. If this is true Ngis added to the temporary set T
and the loop proceeds to its end by selecting all nodes Ng
{Fu
Nk,FNk\Fu
Nk}. The temporary set Tis assigned to be the
new routing table FNkof Nk. Additionally, the LEE counter
#Nkis set to the total number of LEE values of the routing
table #FNkplus one for the node Nkitself. With the next loop
of Algorithm 2, the already linked nodes are removed from all
other nodes in this layer. Finally, to avoid overbooking in the
next layer, one node is removed from the set of backlink nodes
which is able to proceed the link in terms of LEE value to the
next layer towards the BS. This node is selected from the set
of nodes BNkwhich is able to manage at least the link count
of Nk.
The process is repeated for all nodes in set Rnand ends
with a routing table FNkfor each node Nk. In a next step
we move one layer backward in the structure to Rn1and
continue here for all nodes with the just described procedure.
The algorithm navigates backward in the network structure
layer by layer until the BS is reached and all routing tables
are found. The set of all routes FKis returned as set of
all routing tables found for this network Rpunder the given
energy constraints.
A advantageous side effect of this algorithm is the simple
detection of uncovered nodes using Algorithm 3. It can be
easily detected which nodes are simply not within the commu-
nication range and which nodes are within the communication
range but without sufficient link budget. The function GET UN-
COVERED consists of two lines. With the first command these
nodes are found, which are not in communication range URby
subtracting the union of all nodes assigned to any layer form
the complete set of nodes R. The second line starts a recursive
algorithm to determine UL. The recursive algorithm is called
for all forward nodes and with this recursion it visits each
node and adds it up to the linked nodes. And again the linked
nodes are subtracted from Rwhich results in the unlinked
nodes UL. The results therefore can serve as an important
input for rearranging the nodes or adding additional routing
nodes. However, such enhancements are out of the scope of
this paper.
A graphical description of the algorithm is given in Fig. 2,
where the LEE values of nodes Nkare given as integer
numbers. LEE M stands for mains powered and therefore,
not limited by energy. The black lines and circles represent
available communication links and nodes. The green lines
represent resulting routes between BS and end nodes. The bar
colours are used to illustrate the communication range.
III. MEDIA ACCESS AND COMMUNICATION FRAME
ORGANIZATION
Media access is, apart from PHY-layer, the base of the
protocol stack. Therefore, routing is also depending on the
organization and timing of the MAC-layer. None of the avail-
able industrial standards like WirelessHART or ISA100.11a
[20], [21] can sufficiently fulfill all our requirements on the
MAC-layer. Especially, there is no efficient implementation
concept for the required layer structure of the BASEBAL
algorithm. For a high LEE, the energy efficient handshaking is
1
k
20
10
30
21
11
2
3
12
31
13
4
22
23
14
5
BS
R1R2R3R4
R0
5
LEE
1
5
1
2
5
8
1
1
1
1
M
2
2
4
6
Fig. 2: Routing algorithm structure for nodes Nk
Algorithm 3 Get uncovered nodes
1: function GE TUN COV ERED(Rp,FK,R)
2: URR\nmax
S
n=1
Rn
3: ULR\GE TLINKED(Ng,FK)
4: return UR,UK
5: function GE TLINKED(Nk,FK)
6: L=Nk
7: for all NgFNkdo
8: L={L,GE TLINKED(Ng,FK)}
9: return L
taken from the EPhESOS [1] MAC-layer protocol. This MAC-
layer provides also the classical LLC layer from OSI model
with handshaking and flow control. EPhESOS additionally
provides a simultaneous synchronized energy efficient data
streaming from sensors to a centralized BS. This TDMA-based
protocol is optimized for sensor applications and described
in the following section. The network has a common system
clock for synchronizing the measurements, communication and
triggering of events. For energy saving purpose, most of the
time the nodes are in standby or sleep state. For the same
reasons the timing during sleep periods is performed with a
very low clock rate, e.g. 32kHz. Therefore, the system time
granularity is defined by the standby clock and synchronization
is done by periodic beacons sent each SF. Synchronization
and frequency estimation is an integral part of the system and
supported by several concepts presented in [22], [23]. Similar
to WirelessHART or ISA100.11.a, SFs are introduced. One SF
is the time interval between consecutive beacons, which is long
enough to implement the TDMA-scheme of Fig. 3. Each SF
is divided in two parts, the beacon and the time slots for the
Adr SF#T L ACK AdrT L Cmd
N2 N11 N21N1
DATA
SF n
BN31 N99
SF n+1
N2N1
B
AdrT L Data
Beacon Data
2BS
2Node
Fig. 3: Superframe structure
node data. The standard SF has a leading beacon and as many
time slots as nodes are connected to one BS. A SF counter,
SF#, is used as timestamp distributed in the beacon by the BS.
The address in the beacon identifies the node, which is sending
out the beacon. With the address each listening node is able
to identify the time slot of the sender and therefore to align
the local time. Furthermore, the acknowledge information is
contained in the beacon as bit field. Each bit represents the
acknowledge information of one time slot. If the nth bit is set
to 1 the data of the nth time slot was accepted by the receiver.
After this communication and synchronization information,
network commands are added. Here we discuss the routing
relevant commands and leave out the rest of the sensor and
actuator commands. The main purpose of the EPhESOS
Data
Type
Description
T uint8 t type
L uint8 t length
V uint8 t[] T1
L1
V1
T2
L2
V2
Tn
Ln
Vn
TABLE I: Container chunk
representing a datagram
where 1 to nTLV structures
are contained.
Data
Type
Description
T uint8 t type
L uint8 t length
V uint16 t device id
uint32 t superframe num-
ber SF#
TABLE II: Device info data
chunk mandatory in each
EPHESOS datagram
Data
Type
Description
T uint8 t type
L uint8 t length
V uint8 t mode:
overwrite\append
V uint16 t
[]
ID of routed
nodes
TABLE III: Routing data
chunk, as an example of a
flexible payload
protocol family is information exchange under the constraint
of minimizing energy. To reach this goal, any complicated
parsing, redundancies and code mapping has to be avoided. A
very simple scalable datagram structure is used to encapsulate
commands and data. EPhESOS datagram is introduced as
highly flexible and versatile means of communication. Based
on [24], a datagram is a self-contained and independent entity
of data carrying sufficient information to be routed from
the source to the destination. Therefore, EPhESOS datagrams
are byte sequences, called chunks, with a two byte header
representing a Type, Length, Value (TLV) structure. There are
two different basic types of TLV chunks, the container chunk,
cf. Tab. I, which implies other chunks and the individual chunk
containing the information described by the type identifier.
Individual chunks dedicated to one node are placed in one con-
tainer. Thus, one and only one device info chunk is mandatory
in any container chunk. The idea of EPhESOS services and
protocols is dedicated to versatility, so, just few chunk types are
predefined and necessary for routing and other administrative
tasks during configuration. All non-predefined types can be
used as application specific chunks which can be defined
and used to optimally fit the application. Predefined chunks
are chunks for node administration, protocol control, setting,
resetting the routing table and acknowledgment. Tables I to
III describe exemplary a minimal realization of an EPhESOS
datagram and Fig. 3 shows a compilation in a SF. As an
example for leaving out redundancy, we discuss the routing
datagram, cf. Tab. III. The routing datagram is simply the
complete number of nodes assigned to a cluster head which is
the destination of the routing datagram.
A. Routing superframe structure
Although the SF structure is optimized for energy efficient
non-routed communication, we use the same frame structure
also for routed connections to earn the same benefits. In
Fig. 4 such a communication is depicted on MAC-layer with
synchronized SFs. In the following, the time slots assigned
to a node Nxare addressed as T Sx. The BS sends out a
beacon (blue arrow) and node N1is receiving this beacon.
As N1is a routing node it resends the beacon in its assigned
time slot. Additionally, node N1adds its sending data to this
data packet. The BS and N12 are listening in the time slot
of N1, and both devices process the individually addressed
data. N12 is sending its sensor data in its assigned time slot.
During network discovery N1received a routing table, which
assigns T12 to N12. Therefore, N1has to listen to N12 and
buffers the data for sending it to the BS. As depicted in Fig. 4
beacons are processed in the same way as already described.
Additionally, the beacon sent from N1to N12 contains the
ACK to indicate that the last data packet was successfully
received. Furthermore, the data buffered in N1is sent with the
same packet as the N1sensor data. To guarantee the hop delay
from protocol perspective, it is necessary that T S1is long
enough to transmit all routed data. If not, the remaining data
has to be sent within the following SFs. Evidently, the latency
between source nodes and BS is increasing with the number
of layers. Here it can clearly be seen, that also the data length
is contributing to latency and it is necessary to have a defined
data length for a defined latency. If the data length would be
random, we would have to introduce a queuing mechanism in
the routing nodes. This would make the latency random, which
we need to avoid for the considered applications. Besides
that, there is always a non-zero probability to lose packets,
which triggers a packet retransmission. This introduces an
unavoidable statistic part of the latency which cannot be
ignored but minimized as much as possible by reducing hop
N2 N12 N12N1
BS
SF n
N2 N12 N12N1
N1
N2 N12 N12N1
N12
B
B
SF n+1
N2N1
N2N1
N2N1
B
B
t
t
t
N13
N13
N13
hop delay
BB
Fig. 4: Routing example with standard superframe structure
N2 N11 N21N1
BS
SF n
N2 N11 N21N1
N2
N2 N11 N21N1
N11
N2 N11 N21N1
N21
B
B
B
B
N11
N11
N11
N11
N2
N2
N2
N2
SF n+1
N2N1
N2N1
N2N1
N2N1
B
B
B
B
Fig. 5: Routing example with extended superframe structure
distances or increase transmission signal power. A statistical
error analysis is, however, beyond the scope of this paper.
B. Extended routing superframe structure
Due to the demand of latency reduction the standard SF is
modified to an extended superframe (ESF). The extension is
based on a doubling of the time slots assigned to the routing
nodes. The routing node time slots have to be organized by
increasing slot numbers according to the number of the layer
where the node is placed. In Fig. 5 an example for this time
slot organization is given. The time slots of the routing nodes
are T0,T2,T11 where in the EPhESOS protocol T0is as always
occupied by the BS. The time slots of all other routing nodes
are able to send data received in the beacon of the previous
layer. This allows to cascade commands or similar data within
one SF from BS to the last layer of nodes. Additionally, if
the same time slot organization is available on the routes back
from the nodes of the last layer to the BS, the answer of the
far end node can be received within one ESF. This represents
a distinct improvement for the maximum latency, as the round-
trip time is limited to one ESF respectively SF.
IV. ENE RGY CON SU MP TI ON
The routing algorithm described in section II calculates its
routes according to the energy budget of the individual nodes,
which is given in LEE values. It is important to describe the
energy consumption in an abstract and hardware-independent
unit. One LEE is defined as the energy required for the recep-
tion and transmission of one packet. Other operations like, e.g.,
the ramp up of the transceiver is not included in LEE and will
be treated separately. The routing algorithm comprises three
phases for which the energy demand is calculated individually.
A. Discovery phase
In the discovery phase, each node which is listed as
possible routing node, has to switch on its transceiver with
power consumption PRX for two wake up periods (WAPs) and
listens to its neighboring nodes [1]. Additionally, the routing
node has to answer to each detected node at least with one
packet to send the time slot (TS) number assigned to the node
Nkand synchronize the new node. In parallel, the routing node
starts its regular service by sending its beacon, which is also
one packet, with a rate of 1
TSF . Here TSF is the SF duration. The
energy can be bounded by an upper limit given the maximum
number of answering nodes in the network K2, without BS
and the node itself. The energy for one sent packet is EPTX
and for one receive packet is EPRX. For most transceivers these
values are identical. Thus, the energy for a LEE is
ELEE =EPTX +EPRX.(2)
The energy of the discovery phase amounts to
EDisc PRX2TWAP + (K2)EPTX +EPTX
2TWAP
TSF
.(3)
In this phase switching on and off transceivers is not foreseen
and therefore it is not considered.
B. Link phase
During the link phase, the routing tables have to be
transmitted to the routing nodes, first from set R1, then from
set R2and so forth. Due to the fact that the number of routes
through one node is based on its LEE value, its energy is
proportional to LEE including the overhead for transceiver
ramp up EOH. The number of links over one routing node is
calculated by its routing table #Nk= #FNk+1 and therefore,
ELink = #Nk(ELEE +EOH).(4)
C. Operation phase
The energy consumption during the operation phase TOper
is also proportional to #Nk. Additionally the beacon is sent
every SF. Hence, it results in
EOper =TOper #Nk(ELEE +EOH)1
TBuffer
+EPTX
1
TSF .
(5)
The buffering time TBuffer is used to save energy by sending
several sensor data values in one packet. This is because short
packets sent at a high rate increases the overhead due a higher
number of transceiver ramp ups. Moreover, the buffering
appears in the sleep mode in which the power consumption is
usually three orders of magnitude less than power consumption
in duty cycles. The design parameter TBuffer is used to adapt
the network to the application. Additionally, the constraint of
TS or maximum packet length has to be taken in account.
Finally, the three energy values are added up as total energy
demand ETotal to select a power supply
ETotal =EDisc +ELink +EOper.(6)
D. Real world example
The EPhESOS protocol was implemented on different
hardware platforms [25],[1] with similar results on energy
efficiency. Here, an outline is given to show the energy charac-
teristics for an implementation with an nRF51 from NORDIC
Semiconductors, cf. Table IV. Although energy harvesting is
applied [26], [27], here we use a 3V battery CR2032/CR2016
as power source to present a comparable source. The energy
Parameter Value Description
TWAP 5 sec wake up period
TSF 100 msec superframe period
TBuffer 1 sec buffer time
K100 number of nodes
PRX 12mW transceiver power receive mode
PSleep 12µW transceiver power sleep mode
EPTX, EPRX 46µJ packet energy transmit/receive
EOH 36µJ transmit/receive overhead
EBattery 2484J battery energy CR2032
TABLE IV: Electrical system parameters
consumption is given by ETotal for one sensor node without any
sensing device on it. We consider only the communication part.
In the following we illustrate the influence of the LEE value
and the number of links #Nkover one routing node on the
time of operation for one routing node. Hence, we reformulate
(6), (3) and (5) to
TOper =EBattery EDisc #Nk(ELEE +EOH)
#Nk(ELEE +EOH)1
TBuffer +EPTX 1
TSF .(7)
Based on [25], the sleep mode power consumption reduces the
available time of operation by 10% which is neglected here to
show the dependence of TOper on #Nk. Fig. 6 is a convenient
way to select the LEE value based on a given battery and a
desired time of operation.
V. CONCLUSIONS
We presented a communication and routing protocol based
on the EPhESOS protocol and the BASEBAL algorithm to
balance energy consumption among wireless sensor and ac-
tuator nodes under the constraint of minimizing end-to-end
latency. During the discovery phase the network is set up in a
layered structure according to the link quality. This leads to a
minimized latency for the routes from all nodes to the central
BS. The routes passing a layer are distributed among the
routing nodes in this layer according to their available energy
1 4 8 12 16
10
20
30
40
50
60
#N, LEE
TOper
[days]
CR2032 230mAh
CR2016 90mAh
Fig. 6: Operation time TOper of a routing node as function of
the number of routed links for two different batteries.
budget. This budget is derived from an energy calculation that
allows to set the total time of operation. The usage of assigned
time slots for each node in a TDMA-based MAC protocol
leads to a defined latency depending on the hop distance. The
extended SF structure reduces the latency to one SF period.
ACKNOWLEDGMENT
This work has been supported in part by the Austrian Re-
search Promotion Agency (FFG) under grant number 853456
(FASAN: Flexible Autonome Sensorik in industriellen AN-
wendungen) and by research from the SCOTT project. SCOTT
(www.scott-project.eu) has received funding from the Elec-
tronic Component Systems for European Leadership Joint
Undertaking under grant agreement No 737422. This Joint
Undertaking receives support from the European Union’s Hori-
zon 2020 research and innovation programme and Austria,
Spain, Finland, Ireland, Sweden, Germany, Poland, Portugal,
Netherlands, Belgium, Norway.
REFERENCES
[1] H.-P. Bernhard, A. Springer, A. Berger, and P. Priller, “Life cycle of
wireless sensor nodes in industrial environments,” in 13th IEEE In-
ternational Workshop on Factory Communication Systems, Trondheim,
Norway, May 2017, pp. 1–9.
[2] N. A. Pantazis, S. A. Nikolidakis, and D. D. Vergados, “Energy-
efficient routing protocols in wireless sensor networks: A survey,” IEEE
Communications Surveys Tutorials, vol. 15, no. 2, pp. 551–591, Second
2013.
[3] H. H. Liu, J. J. Su, and C. F. Chou, “On energy-efficient straight-line
routing protocol for wireless sensor networks,” IEEE Systems Journal,
vol. 11, no. 4, pp. 2374–2382, Dec 2017.
[4] S. Li, S. Zhao, X. Wang, K. Zhang, and L. Li, “Adaptive and secure
load-balancing routing protocol for service-oriented wireless sensor
networks,” IEEE Systems Journal, vol. 8, no. 3, pp. 858–867, Sept
2014.
[5] K. Samara and H. Hosseini, “A routing protocol for wireless sensor
networks with load balancing,” in 2016 IEEE International Conference
on Electro Information Technology (EIT), May 2016, pp. 156–161.
[6] T. T. Huynh, A. V. Dinh-Duc, and C. H. Tran, “Delay-constrained
energy-efficient cluster-based multi-hop routing in wireless sensor net-
works,” Journal of Communications and Networks, vol. 18, no. 4, pp.
580–588, Aug 2016.
[7] T. T. Huynh, T. N. Tran, C. H. Tran, and A. V. Dinh-Duc, “Delay
constraint energy-efficient routing based on lagrange relaxation in
wireless sensor networks,” IET Wireless Sensor Systems, vol. 7, no. 5,
pp. 138–145, 2017.
[8] E. Bakin, G. Evseev, and D. Dorum, “Load balancing routing algorithm
for data gathering sensor network,” in 2012 12th Conference of Open
Innovations Association (FRUCT), Nov 2012, pp. 1–6.
[9] A. E. Fawzy, M. Shokair, and W. Saad, “Balanced and energy-efficient
multi-hop techniques for routing in wireless sensor networks,” IET
Networks, vol. 7, no. 1, pp. 33–43, 2018.
[10] W. R. Heinzelman, A. Chandrakasan, and H. Balakrishnan, “Energy-
efficient communication protocol for wireless microsensor networks,
in Proceedings of the 33rd Annual Hawaii International Conference on
System Sciences, Jan 2000, pp. 10 pp. vol.2–.
[11] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, “An
application-specific protocol architecture for wireless microsensor net-
works,” IEEE Transactions on Wireless Communications, vol. 1, no. 4,
pp. 660–670, Oct 2002.
[12] P. Li, S. Xu, K. Sun, X. Qiu, and F. Qi, “A path planning method of
wireless sensor networks based on service priority,” in 2017 13th In-
ternational Conference on Network and Service Management (CNSM),
Nov 2017, pp. 1–4.
[13] “IEEE Standard for Local and metropolitan area networks: Media
Access Control (MAC) Bridges,IEEE Std 802.1D-2004 (Revision of
IEEE Std 802.1D-1998), pp. 1–277, June 2004.
[14] T. Gomes, L. Guardalben, P. Salvador, and S. Sargento, “A spanning
tree protocol over mobile wireless ad hoc networks,” in 2013 IEEE
International Conference on Communications Workshops (ICC), June
2013, pp. 1089–1094.
[15] D. Allan, J. Farkas, and S. Mansfield, “Intelligent load balancing for
shortest path bridging,” IEEE Communications Magazine, vol. 50, no. 7,
pp. 163–167, July 2012.
[16] J. So and H. Byun, “Load-balanced opportunistic routing for duty-
cycled wireless sensor networks,IEEE Transactions on Mobile Com-
puting, vol. 16, no. 7, pp. 1940–1955, July 2017.
[17] L. Chen, J. Warner, W. Heinzelman, and I. Demirkol, “Mh-reach-
mote: Supporting multi-hop passive radio wake-up for wireless sensor
networks,” in 2015 IEEE International Conference on Communications
(ICC), June 2015, pp. 6512–6518.
[18] Y. Liu, Q. Chen, G. Liu, H. Liu, and Q. Yang, “Ecosense: A hard-
ware approach to on-demand sensing in the internet of things,” IEEE
Communications Magazine, vol. 54, no. 12, pp. 37–43, December 2016.
[19] I. S. Lima and L. C. Kretly, “A new approach on wake-up radio
(wur) based on active rfid,” in 2017 IEEE International Conference
on Microwaves, Antennas, Communications and Electronic Systems
(COMCAS), Nov 2017, pp. 1–3.
[20] S. Petersen and S. Carlsen, “WirelessHART Versus ISA100.11a: The
Format War Hits the Factory Floor,” IEEE Industrial Electronics
Magazine, vol. 5, no. 4, pp. 23–34, Dec 2011.
[21] O. Khader and A. Willig, “An energy consumption analysis
of the wireless hart tdma protocol,” Computer Communications,
vol. 36, no. 7, pp. 804 – 816, 2013. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0140366413000029
[22] H. P. Bernhard, A. Berger, and A. Springer, “Timing synchronization
of low power wireless sensor nodes with largely differing clock fre-
quencies and variable synchronization intervals,” in 2015 IEEE 20th
Conference on Emerging Technologies Factory Automation (ETFA),
Luxemburg, Luxemburg, Sept 2015, pp. 1–7.
[23] H.-P. Bernhard and A. Springer, “Linear complex iterative frequency
estimation of sparse and non-sparse pulse and point processes,” in 2017
25th European Signal Processing Conference (EUSIPCO 2017), Kos,
Greece, Aug. 2017, pp. 1150–1154.
[24] A. Tanenbaum and D. Wetherall, Computer Networks, ser. Pearson
custom library. Pearson, 2013.
[25] A. Berger, A. P¨
otsch, and A. Springer, “Tdma proposals for wireless
sensor networks for highly reliable and energy efficient data collection
in an industrial application,” in Proceedings of 2012 IEEE 17th In-
ternational Conference on Emerging Technologies Factory Automation
(ETFA 2012), Sept 2012, pp. 1–8.
[26] A. Berger, T. H ¨
olzl, L. B. H¨
ormann, H. P. Bernhard, A. Springer,
and P. Priller, “An environmentally powered wireless sensor node
for high precision temperature measurements,” in 2017 IEEE Sensors
Applications Symposium (SAS), March 2017, pp. 1–6.
[27] L. B. H¨
ormann, A. Berger, A. P¨
otsch, P. Priller, and A. Springer,
“Estimation of the harvestable power on wireless sensor nodes,” in 2015
IEEE International Workshop on Measurements Networking (M N), Oct
2015, pp. 1–6.
... More detailed examination of modern heterogeneous WSNs based usually on IEEE 802.15 standard that are used for IIoT solutions is required due to the limited hardware capabilities (simple low-cost end devices), much higher network density and demanding propagation environment. To address all the problems mentioned above, a novel approach that combines the concept of network layering presented in our work [37] and sensor nodes equipped with the ESPAR antenna [38], [39], [40] has been proposed in this paper. In [37], a routing algorithm was presented that optimizes the energy consumption and minimizes the latency by searching an optimal routing scheme. ...
... To address all the problems mentioned above, a novel approach that combines the concept of network layering presented in our work [37] and sensor nodes equipped with the ESPAR antenna [38], [39], [40] has been proposed in this paper. In [37], a routing algorithm was presented that optimizes the energy consumption and minimizes the latency by searching an optimal routing scheme. In this work, we leave out the routing aspects and focus on finding an optimal layered network structure by introducing relay nodes equipped with ESPAR antennas and are capable to change the radiation pattern based on the network requirements [40]. ...
... In this work, we leave out the routing aspects and focus on finding an optimal layered network structure by introducing relay nodes equipped with ESPAR antennas and are capable to change the radiation pattern based on the network requirements [40]. Differently from [37], we consider the link QoS to discover the network structure. By the boost of directional beams and the relaying, a stringent link QoS can be achieved. ...
Article
Full-text available
Industrial Internet of Things (IIoT) applications require reliable and efficient wireless communication. Assuming dense Wireless Sensor Networks (WSNs) operating in a harsh environment, a concept of a Time Division Multiple Access (TDMA) based WSN enriched with Electronically Steerable Parasitic Array Radiator (ESPAR) antennas is proposed and examined in this work. The utilized antenna provides one omnidirectional and 12 directional radiation patterns that can be electronically switched by the sensor node. We introduce a relay discovery algorithm which selects those sensor nodes with an ESPAR antenna capable to act as relay. The selection of the relay nodes is based on a certain link quality threshold that algorithm uses as input. The outcome is a reduction in number of layers or hops with a guaranteed Quality of Service (QoS). To emphasize the physical aspect of the wireless propagation, we introduce the measured antenna radiation patterns and consider two different path loss propagation models representing blockage-free and blockage-prone industrial environments. A number of network simulations were performed and Signal-to-Noise Ratio (SNR) as a link quality measure was examined with respect to the network density and different measured radiation pattern settings. The main outcomes show a trade-off between SNR per link and the percentage of nodes that can serve as relays. As a result, we propose network design guidelines that take under consideration the QoS range with respect to SNR together with an optimal number of antenna radiation patterns that should be selected as a trade-off between latency, energy consumption and reliability in a network.
... Other approaches like [17] and [18] consider heuristic functions to balance the routing among the relaying nodes. A routing algorithm that minimizes the network latency and balances the energy consumption of the nodes has been introduced in [19]. In comparison to the approach of low-energy adaptive clustering hierarchy (LEACH) algorithm [20], [21], where the wireless network is clustered to achieve energy optimization with randomly selected cluster heads, authors in [19] do not consider random cluster heads. ...
... A routing algorithm that minimizes the network latency and balances the energy consumption of the nodes has been introduced in [19]. In comparison to the approach of low-energy adaptive clustering hierarchy (LEACH) algorithm [20], [21], where the wireless network is clustered to achieve energy optimization with randomly selected cluster heads, authors in [19] do not consider random cluster heads. The route metrics are compared from an energy perspective only within equal hop distances. ...
... In contrast to most other methods, [19] introduces a routing algorithm that integrates the MAC layer as part of the routing. Hence, it is possible to apply a synchronized TDMA scheme in more than one layer so that the overall synchronization is guaranteed. ...
Article
Full-text available
Minimized and nearly deterministic end-to-end latency facilitates real-time data acquisition and actuator control. In addition, defined latency is an integral part of quality oriented service in order to get closer to the reliability of wired networks and at the same time take advantage of wireless networking. This paper introduces a QoS routing protocol capable of balancing power consumption between wireless sensor and actuator nodes while minimizing end-to-end latency. We introduce a TDMA scheme in the routed wireless network to enable defined latency and in addition it improves the energy efficiency by avoiding collisions which eliminates time and energy consuming retries. Our novel routing method allows latency and round-trip times to be calculated in advance. We implemented a demonstrator and show experimental results of a wireless sensor network with our proposed routing scheme.
... Trust indicators and configurability supporting privacy labeling was part of the system design from the start. Other trust factors like system availability are addressed in technology building blocks balancing sustainable energy management and dependable connectivity with appropriate new protocol approaches like [12]. Project SCOTT will display and report results in a multitude of demonstrators at its closure event in 2020; activities and progress are reported frequently via several channels 19 . ...
Preprint
Full-text available
Digitalization is proving to be a game changer in bridging the gap between the heterogeneous skills and markets. It increases productivity through optimisation over the entire supply chain and lets new services emerge through the convergence of applications domains. In this paper, we are providing a review of the main automotive trends and are highlighting how digitalization (especially by information communication technologies-ICT) is supporting, even pushing innovation. We are especially mapping to the IOT4CPS and SCOTT projects to present key results related to Internet of Things supporting the digital transition in the automotive domain.
Article
Full-text available
In Wireless Sensor Networks (WSNs), the power resources of the nodes are significantly restricted. Hence, a special treatment for their available energy is deeply required. In long distance transmissions, Multi-Hop (MH) techniques are preferred as they minimize the amount of energy cost consumed by each node along the path. No doubt that finding the optimal routing path between nodes is one of the most interesting issues in MH. This paper proposes a Balanced and Energy Efficient Multi-Hop (BEEMH) algorithm that is developed based on Dijkstra algorithm. BEEMH gives great interest to the residual energy of nodes, hence higher energy nodes are exclusively elected to work as relays. Moreover, the total energy consumption at both TX and RX is merged to represent the weight of links between nodes. After that, the concept of Dijkstra algorithm is employed to efficiently search for the minimum cost path. Thereafter, two proposed protocols, namely, BEERAD and BEERAD-E are introduced. Both are centralized protocols, mainly based on the BEEMH algorithm. MATLAB simulator has been used to evaluate BEEMH algorithm in comparison with other conventional algorithms such as; MTE, ESLHA, and ESRAD under various scenarios of network models. Then, the performance of our proposed protocols is compared with the related MH protocols. From the Simulation results, we conclude that both BEERAD and BEERAD-E are more balanced, reliable and energy efficient in comparison with other related protocols. Hence, the lifetime and the stability period of the network is improved by a large degree. Also, the total throughput is increased.
Conference Paper
Full-text available
We present the design of a suite of protocols for wireless sensor networks (WSNs) with respect to a complete life cycle of a WSN node from warehouse to the end of operation. While there are numerous publications on various, usually isolated, aspects of WSNs, the whole life cycle of a node from registration in an automation system via warehouse, calibration, mounting, performing measurements to finally unmounting, has not yet been sufficiently addressed as compound survey. Our application example is a WSN to be used in automotive test beds in which a large amount of testing with many different sensors is performed in controlled environments. While there is published work on WSNs for performing the measurements focusing on node hardware and MAC protocol, we now extend this work by accounting for the whole life cycle of operation of such a WSN and its nodes. This is mainly achieved by introducing optimized MAC protocols for wireless communication in all life cycle phases. Right from beginning of the life cycle the nodes are synchronized with a base node. Even during long offline periods nodes stay synchronized. The life cycle is modeled via a set of states, instantiated in state machines, which control operation in the base station and the nodes. Besides, considering the whole life cycle of the sensor nodes, our design minimizes energy consumption, largely avoids collisions due to suitable multiple access protocols, and allows tight synchronization even during long sleep periods. A demonstrator concludes the presentation and shows functionality and benefits of the concept.
Article
Full-text available
An Internet of Things system typically contains a large number of low-cost devices that are mainly powered by batteries and designed to operate for a relatively long period of time. Due to the stagnated battery technology, energy efficiency will still be a burning issue for future IoT systems. To achieve better energy efficiency, we propose an innovative sensor architecture called EcoSense. Unlike traditional software-based approaches (e.g., duty cycling or adaptive sampling techniques), EcoSense provides a hardware-based on-demand sensing mechanism that effectively eliminates the energy waste caused by a sensor working in sleep or standby mode. When desired events are not present, an EcoSense sensor is completely turned off to save energy (i.e., drawing zero current). When desired events occur, an on-demand connection module will harvest energy from the events and reactivate the sensor. We implemented light- and RF-driven EcoSense sensors, and evaluated their performance in realworld experiments. Experimental results show that the EcoSense technique is able to immediately detect lights and RF signals, and offers reasonable reaction distances.
Article
Most of researches on wireless sensor network (WSN) are focused on how to reduce energy consumption to increase the network lifetime. There are few researches solving the problem of energy optimization ensuring delay constraint. Whereas, the delay is important factor for applications that require delay-sensitive data. Several works have been published to balance the energy consumption and data delay. They achieved many different results but each proposal has certain limitations. Some proposals have high computational and messaging complexity. Some others have not found the optimal solution. In this study, we investigate in finding solution to improve the energy efficiency of sensor nodes satisfying end-to-end delay to transmit data from sensor nodes to sink in multi-hop WSNs. Based on Lagrange relaxation method, we propose an aggregate cost function between energy consumption and delay as well as an efficient method to find the optimal multiplier for that objective function. We provide two algorithms to find paths with least energy consumption while maintaining end-to-end delay requirement from any sensor node to sink. Besides analyzing the complexity and convergence of the algorithm, the simulation results also show that the proposed algorithm achieved good balance between energy consumption and delay compared with the previous proposals.