Click here to load reader
Upload
trinhthu
View
214
Download
2
Embed Size (px)
Citation preview
A Wireless Mesh Communication Protocolfor Smart-metering
Daniel GeelenEindhoven University of Technology
Netherlands
Gert van Kempen &Frans van Hoogstraten
SmartDutch
Netherlands
prof.dr. Antonio LiottaEindhoven University of Technology
Netherlands
Abstract—Worldwide there has been increasinginterest over the past few years for so-called “SmartMeters”, in academia, governments and in industry.Such smart-metering systems need a way to commu-nicate the collected data reliably and cost efficientlyto the back-office for analysis. Several competingtechnologies exist and are in use world-wide. Mesh-networks have been the winning technology in theUSA and Australia for the past two years, and aregaining interest in Europe at the moment due to itsreduced costs and increased reliability. In this paperwe present and evaluate a real-life implementationof a new routing protocol for use in smart-meteringmesh-network grids. The routing protocol we presentis designed with both technological constraints andlegislative requirements as posed by the applicationarea in mind. Our evaluation of the protocol is basedon real-world experience and data collected fromreal-world devices, in combination with simulationstudies of the protocol. Our evaluation shows that theprotocol is a robust, reliable solution for communicat-ing collected data in difficult scenarios, showing greatresilience against both bit-errors and node-failures.
I. INTRODUCTION
In this paper we present (section III, IV) and
evaluate (section V, VI) a new wireless mesh
protocol for use in residential smart-metering appli-
cations. This protocol has been under development
by a Dutch company and has already been success-
fully deployed in case-study installations commis-
sioned by three out of the four largest Dutch grid
operators. In these case-study installations the pro-
tocol appeared to work very well, reaching close to
100% delivery rates. Our simulations confirm these
results and show that the protocol can be expected
to maintain these high delivery rates up-to bit-error
rates as high as 10−4. We use data collected from
these case-studies to analyse the performance of
the protocol, as well as to aid in verification and
validation of our simulations. Because the protocol
is targeted for deployment in households across
the European Union it is subject to European law
(section II). This combined with the constraints
dictated by the available hardware platform make it
challenging to design an efficient, reliable protocol.
We present the protocol in section III.
Smart-metering refers to the next-generation of
(mostly) electricity meters which need not be re-
stricted to measuring only electricity usage; these
can be combined with gas, water, and heat-meters.
Smart-meters can be differentiated from other mod-
ern metering techniques such as automated meter
reading (AMR) and automatic meter management
(AMM) by the amount of control smart-metering
gives over the monitored devices [1]. Smart-meters
are able to manage connected devices with short
feedback and response times, and offer several
advantages over the traditional metering-devices
that until now have been in use for both utility com-
panies and consumers alike [2]. Because it is pos-
sible for suppliers to interact quickly with smart-
meters, it is possible to offer greater tariff variety
and flexibility in pricing. Smart-meters can also
be remotely administered, making it is easier for
consumers to change suppliers. This in turn leads
to increased competition among suppliers. Con-
sumers also gain more accurate meter reading (and
hence billing). However these same features benefit
suppliers as well. In the same way consumers
find it easier to change suppliers, on the supplier
side the process also becomes more streamlined
and less involved. Metering-device (de)activation
and maintenance can be handled remotely, cutting
down on costs.
Radio Frequency (RF) communications present a
reliable, cost-effective technology for use in smart-
metering networks. Other possible approaches such
as power-line communications (PLC), GPRS and
GSM based access-points or the use of Telephone
(land) lines and (coaxial) cabling suffer limitations
on reliability, bandwidth and costs [1], [3], [4].
A mesh-type network is the natural topology for
a RF-communications based smart-metering net-
work. Solutions exist based on standards such as
IEEE 802.15.4 [5] or ZigBee [6], as well as many
Workshop on Computing, Networking and Communications
978-1-4673-0009-4/12/$26.00 ©2012 IEEE 343
proprietary solutions. Routing-protocols used in
smart-metering solutions range from very simple
ALOHA-like approaches like [7], through the clas-
sical mesh routing protocols [8] as used in ZigBee
or IP-based networks such as 6LoWPAN [9], to
intricately complex solutions such as [10], which
details a project similar to the work presented in
this paper. The system described in [9] has also
been successfully field tested, and uses a TDMA
protocol with polling initiated by the central node
to query metering nodes. It relies on highly precise
time synchronisation between nodes, to be imple-
mented in a customised physical layer, to achieve
collaborative transmission [11], [12]. Although the
protocol described in this paper also requires time
synchronisation, it can use timers with regular
accuracy. Further it is designed to operate within
more strict constraints with respect to bandwidth
use, transmission times and hardware restrictions,
using readily available components.
II. CONSTRAINTS AND REQUIREMENTS
In terms of regulatory constraints the protocol
must take the intended deployment area into con-
sideration. In Europe there are efforts underway to
standardise smart-metering interfaces on the com-
munity level (Mandate CEN/CENELEC M/441), as
well as on the national level (such as the Dutch
NTA-8130 directive or the DSMR (Dutch Smart
Meter Requirements) document [13]) The general
objective of these efforts is to develop standards to
enable interoperability of utility meters. Not only
should the protocol adhere to these interoperability
guidelines, but because the metering-platform and
its proposed protocol will employ wireless mesh-
networking technology restrictions on the use of
radio frequencies have to be taken into consider-
ation also. On the technological level limitations
are imposed by what is possible on a given hard-
ware platform. For the case-studies the hardware
platform used was a ‘Nordic RF 905’. This RF
transmitter chip can transmit in one of 433MHz,
868MHz or 915MHz bands. Due to regulations
and concerns of the development team about the
reliability of the 433MHz band the 868MHz band
was used. On the legislative level1 restrictions may
be placed on the use of certain frequency bands
in terms of Effective Radiating Power (ERP), the
amount of bandwidth used, and the amount of
time a device is allowed to actively transmit (the
active duty cycle). Concretely this means that for
1In the European Union there are regulations which affectall union members. The ECO Frequency Information Systemdatabase (http://www.efis.dk/) provides a reference.
the chosen 868MHz band that the protocol has
to ensure that none of the participating metering-
devices exceeds a 1% duty cycle, at a maximum
ERP of 25mW. The protocol should enable two-
way communication with the utility’s back-office in
order to facilitate the reporting of energy, gas, water
and reverse-energy readings from the different me-
ters to the utility’s back-office, as well as sending
various notifications to the meters. The protocol
should be self-configuring as much as possible,
and should not require expert knowledge to setup.
Minimal preparations for each device (for example
setting an identifier for each device) should be
performed by the fabricator.
III. PROTOCOL DESCRIPTION
In the protocol we distinguish two types of
nodes, metering nodes and concentrator nodes.
Metering nodes are nodes which are attached to
a gas, water, heat, or electrical meter. These nodes
take measurements and can be considered to be
the network’s source-nodes. Concentrator nodes are
nodes which are connected to an Internet connec-
tion, for example through a GPRS up-link. Concen-
trator nodes collect the measurements for transmis-
sion to the back-office, and can be considered as
the sink-nodes in the network. Nodes are organised
in this manner since typically each metering node
only needs to report a small amount of information
to the back-office. By having a single concentrator
node there is only one shared Internet connection
required for a large set of nodes. In this way it
is possible to cut costs compared to the traditional
approach where each node has its own GPRS link.
The protocol tries to keep as much intelligence
in the source-nodes as possible. This means that
there is no polling of the nodes done by the concen-
trator, in contrast to for example [10]. Instead the
metering-nodes themselves initiate communication
on certain pre-set times, and communication takes
only a small amount of time. That is to say each
node has its own communication slot. When nodes
are installed no special attention is paid to any
relation between their node-id and their physical
positioning. This means that even in the event that
two nodes lose time synchronisation, it is unlikely
that their communications will interfere, since they
are likely to be in different areas of the network.
The protocol uses a proprietary packet format
which is optimised for use in meshed communi-
cations among a limited number of nodes, each
of which requires only a small amount of data
to be delivered. Each packet is exactly 32 bytes
long, regardless of whether or not those bytes are
344
actually fully utilised2. The protocol imposes a
15 byte overhead, leaving 17 bytes free for data
communications. Besides source and destination
IDs the header offers, among others, distinction
of packet-type, space for a list of intermediate-
nodes (the packet’s route), and a set of boolean-
flags indicating which of these intermediate nodes
have already forwarded the packet. In comparison,
a raw 802.15.4 header is at least 9 bytes3, but
at that size does not offer distinction of packet
type, hop limitation, or routing information. When
6LoWPAN is used this increases the size of the
header to 16 bytes [9]. Although offering routing
capabilities, the complexity of 6LoWPAN is much
greater, and putting more strain on both imple-
menters and hardware.
Nodes do not keep a routing table in the tra-
ditional sense, but they do cache previously es-
tablished communication-routes. The protocol op-
erates by first inspecting a node’s route-cache to
determine if the route knows a route to the desti-
nation node. If a route is found in the cache the
node will attempt to communicate via this route,
otherwise the node will attempt to communicate
directly. If this fails the node initiates a search-
phase in which it attempts to locate the destination
node at increasing hop distances. Upon receiving
an acknowledgement to its search the node caches
the discovered route and uses this for future com-
munications.
The protocol defines just a few packet-types;
two types of communication packet and one search
packet. Each of packet-type has a corresponding
acknowledgement packet. Figure 1 shows how each
packet type is handled by the nodes. The first
type of communication packet is used when there
is no need to use meshing for communication.
This is called direct communication. In the di-
rect communication mode a node will attempt to
communicate with its intended destination simply
by sending a direct communication packet with
the proper destination ID set, and waiting up-to
250ms for an acknowledgement. When a node
cannot communicate directly with its target node
(e.g. it is out-of-range, or the sending node fails
to receive a timely acknowledgement), meshed
communication is used. The route which is used for
the meshed communication is discovered through a
controlled flooding search. The search is initiated
2This is due to a limitation of the RF-receiver, which requiresthat packet lengths are known before reception in order toperform on-chip CRC verification.
3http://lia.deis.unibo.it/research/WANDA/architecture/amipv6.html#ipv6
by broadcasting a query message containing the
ID of the target node, as well as a limit on the
number of retransmissions onto the network. Any
node which receives the broadcast (and is not the
destination node) verifies that
• it did not initiate the search itself,
• it has not forwarded this search before (this
can be determined by inspecting the packet’s
route-list)
• propagating the search does not cause the
search to exceed the specified hop-limit,
• it is not already in the process of forwarding
another packet,
If all conditions hold the node records its ID in
the first free position in the packet’s route-list, and
selects a random delay between 0 and 56ms (in
8ms increments) and waits this amount of time
before re-broadcasting the search. By including a
small random delay between forwarded messages
the protocol avoids excessive collisions among
nodes attempting to forward search packets. When
the search packet is received by the destination
node, it waits a little over 63msבretransmission-
limit’ before sending an acknowledgement packet.
Because the retransmission-limit for each packet
is set at most 5, this amount of time is sufficient
to ensure that all other nodes have finished their
retransmissions of the search-packet.
Sending the acknowledgement to the search
back to the initiating node happens using meshed-
communication in the same way as the normal
mesh-packet type is meshed over the network.
Upon receiving a packet a node checks whether
it should forward this packet or not. A packet is
eligible for forwarding if it has the correct type
(e.g. an acknowledgement to a search or a mesh-
type packet) and has not yet been forwarded by
this node. Additionally only one packet may be
forwarded at a time. The packet is then prepared for
forwarding by flipping the flag-bit corresponding to
the node’s ID in the packet’s route. The packet is
then retransmitted after a short, random delay (this
delay is calculated in the same way as the delay
used during the search-phase). In this way meshing
a packet in this protocol is a very simple operation
requiring only a few simple checks and a single
bit-flip before the packet is ready for forwarding.
The protocol thus offers a highly efficient mech-
anism for forwarding packets, but does not specif-
ically keep track of the durations of transmissions.
It relies on keeping messages as short as possible
and traffic as low as possible to avoid exceeding the
1% duty cycle limit. Although this approach the-
345
receivepacket
destinationreached?
is DIR?
is DIR?
is SCH?
is SCH?
is MESH?
is MESH?
processpacket
wait
processpacket
wait
discardpacket
forwardpacket
wait
processpacket
processpacket
wait
sendacknowl-edgement
processpacket
yes
no
yes
no
yes
no
yes
no
yes
no
yes
no
yes
no
Figure 1: Decision process upon receiving a packet.
oretically allows for individual devices to violate
the stated requirements, we will show in section V
that even under heavy load the protocol operates
well within the 1% duty cycle limit.
IV. SIMULATION IMPLEMENTATION
For the purpose of evaluating the performance of
the protocol we have created an implementation of
the protocol in the OMNeT++4 network-simulator.
This implementation is based on the existing Cimplementation which runs on the real nodes in
the case study.
For our implementation we make use of the
MiXiM5 framework, which provides radio wave
propagation models, interference estimation, MAC
protocols and more. Instead of only using MiXiMlike a traditional library, providing ready-made
functionality, we have also used several of MiXiM’s
‘simple’ models as a base for the implementation
of our own models. For example we have ex-
tended MiXiM’s SimplePathlossModel with our
own obstacle model. Although an obstacle model
4http://omnetpp.org/5http://mixim.sourceforge.net/
is advertised in [14], it is at present not yet avail-
able in MiXiM. In our obstacle model arbitrarily
shaped and rotated obstacles can be placed in the
simulation. The signal’s attenuation is based both
on the signal passing through the walls (exterior)
of the obstacle, and on the distance the signal
travels through the obstacle. Our obstacle model
is similar in capabilities to the obstacle model
currently being developed6 for MiXiM but differs on
the implementation level, offering more flexibility
in defining the shape of obstacles.
Using a packet-sniffer operational data was ex-
tracted from the case-study installation, and used to
verify the accuracy of the simulation in two ways.
Firstly we created a tool to visualise both the output
of the simulation, as well as the data from the
case-study. Figure 2 is generated using data of a
simulation run, and shows various routes as used
by nodes in the network. Metering nodes are shown
as circles, the trapezium represents a concentrator
node. The thickness of the arrows indicates the
amount of traffic between nodes. As can be seen the
network is more or less split into a set of commu-
nicating nodes above and below the concentrator
node. This output looks very similar to the same
visualisation generated using data from the case-
study installation. The cause of this is a strip of
trees and shrubs present in between these nodes,
blocking the wireless signal. The main reason for
developing our obstacle model is to simulate the
signal attenuation caused by this foliage.
Secondly we have created a similarity index
for comparing the routes used by node in the
real-world versus those generated by the simula-
tion. We have based our approach on Levenshtein-
distance, but have modified it such that it takes
into consideration that routes are generally short,
irregular and use nearby nodes. This modification
was necessary because using an approach based
purely on string edit distance neglects the fact
that nodes are more likely to transmit packets to
nodes which are geographically closer. A detailed
explanation of the algorithm is omitted here due
to space limitations, but we give pseudo-code of
the algorithm used to compare two routes (lists of
nodes) A, B in listing 1. Here function δ(ν1,ν2)gives the distance between two nodes ν1 and ν2.
Using our similarity index we are able to show
that the simulated nodes are discovering and using
routes which are relatively similar (with respect to
the random nature of the protocol) to those used
6http://www7.informatik.uni-erlangen.de/∼sommer/omnet/obstacles/
346
by the real-life nodes.
|[ var A,B : P → ν ; δ ∈ ν×ν→ R ;
i, j,a,b ∈ N ; s,score, la, lb ∈ R
| score := 0
la, lb = length(A), length(B)
if la > 3 ∧ lb > 3 →i := 0
do i < 7 →s := ∞j := 0
do j < 7 →a := �i · (la÷7)+(1÷ (2 · la))�b := �i · (lb÷7)+(1÷ (2 · lb))�if b > 0 ∧ b < lb →
s := s ↓ δ(A[a],B[b−1])2 ↓δ(A[a],B[b]) ↓δ(A[a],B[b+1])2
fij := j+1
odscore := score + si := i+1
odfireturn score
]|Listing 1: Algorithm to calculate a similarity index
for two paths A, B.
V. EXPERIMENTAL RESULTS
Through our simulation we are able to collect
many statistics for each experiment, ranging from
the per-node active duty cycle, through the number
of successfully delivered packets, to the distances
packets travel through the network. We first present
some of the results for an experiment where we
have investigated the performance of the protocol
under a variety of bit-error rates. In our simulations
we assume that the hardware is reliable and any
bit-errors are due to effects in the transmission
medium (e.g. interference, packet collisions, etc.).
Recall that the hardware performs a CRC-check
on all received packets and that all packets are of
equal length. Therefor this also tests the protocol’s
ability to cope with packet-losses, because a spe-
cific bit-error rate directly equates to an expected
packet-loss rate. We also take a brief look at
the reliability of the protocol when faced with a
reduced operating capacity in terms of the number
of available nodes. In this experiment we varied
the number of nodes that remain in operation, both
themselves attempting to use the mesh-network for
communication as well as operating as intermediate
nodes in the network.
Figure 3 shows that the protocol’s ability to
deliver measurements to the concentrator node
is largely unaffected by bit-error rate. Only at
high bit-error rates the packet loss rate starts to
negatively affect the protocols ability to reliably
deliver packets. However even at a bit-error rate of
10−4 the protocol still delivers more than 99.5%
of measurements to the concentrator. Here it can
also be seen that even though a node may itself
believe to have failed to deliver the measurement
due to not receiving an acknowledgement, even
at very high bit-error rates of 5.5 ·10−4 (expected
packet loss rate > 15.8%), still more than 90% of
all measurements are delivered at the concentrator.
Figure 4 shows three measures for the length
of the routes. Firstly it shows the average hop-to-
hop transmission distance and the average distance
covered by a route. These are measured in meters
(left axis). Secondly it shows the average number
of hops a route uses (right axis). Similarly to
Figure 3, Figure 4 shows that routes are fairly
unaffected by changes in the bit-error rate. Again
only at very high bit-error rates a change in the
length of routes can be observed.
Figure 5 shows the average duty cycles for both
the concentrator node and for the metering nodes.
The concentrator node has to handle relatively large
amounts of traffic, because every other node will
be communicating with it to report measurements.
The figure shows that the protocol operates well
under the imposed 1% duty cycle imposed by
regulations, even at high bit-error rates where many
re-transmissions may be needed in order to de-
liver the measurement to the concentrator node. It
can be seen that at the breaking point (compare
Figure 3) the upward trend for the concentrator
node’s duty cycle is broken. This is because at
this point the channel state is so bad that commu-
nication becomes effectively impossible for nodes
which require more than a few hops to reach the
concentrator node, thus alleviating stress on the
concentrator node.
The second experiment, where we varied the
number of nodes that remain in operation to simu-
late nodes failing over time, shows that the protocol
is also able to handle large losses in operating ca-
pacity well. For this experiment we have assumed
a bit-error rate of 3.3 · 10−05, the bit-error rate
at which, as the previous experiment showed, the
protocol starts to become affected by packet losses.
Figure 6 shows that at 50% capacity, i.e. only half
347
of the nodes remain available for meshing, we are
still able to deliver over 95% of the measurements.
Since by the previous experiment we know that
at the given bit-error rate the protocol is able to
deliver close to 100% of the measurements, we
can infer that for this particular scenario, even
if half of the nodes have failed, there is still a
very high chance that 95% of the remaining nodes
will be unaffected. The large variance is caused
by the fact that if too many nodes around the
concentrator node fail, the network becomes dis-
connected from the concentrator node, and none of
the nodes can deliver its measurements. Estimating
the probability with which the network disconnects
is hard, and would require more simulation runs to
be performed.
VI. CONCLUSIONS AND FUTURE WORK
We have described various requirements
to which smart-metering protocols targeting
a widespread deployment should take into
consideration. We have touched on legislative,
technological and economic considerations which
affect a protocol’s design. We have presented a
new proposed protocol for use in smart-metering
applications, designed to operate under the given
constraints. We have simulated the protocol under
a variety of circumstances representing possible
real-world situations and show that the protocol
is a robust, reliable solution for communicating
collected data even when faced with high bit-
error rates or failure of many nodes in the
network. Bit-error rates have little effect on the
protocol’s reliability or communication latency.
The protocol is robust against node failure, and
ensures that working nodes route around failed
nodes effectively even at relatively high node
failure rates. We show that the protocol stays well
below the stated 1% duty cycle limit imposed
by European law, even at peak retransmission
rates. This leaves room for sending more frequent
updates or adding new features.
Future work includes performing further experi-
ments to assess the protocol’s capacity. This could
be tested by increasing the number measurements
to be delivered, or reducing some of the delays
involved in transmissions. Also it would be inter-
esting to see how the protocol deals with other
topologies. The current scenario consists of a re-
creation of a real-world sub-urban neighbourhood.
Other scenarios could be high-rise constructions
or sparsely populated areas, representing different
challenges each. Results on some of these experi-
ments are presented in [15].
REFERENCES
[1] G. Deconinck, “An evaluation of two-way communicationmeans for advanced metering in Flanders (Belgium),”2008 IEEE Instrumentation and MeasurementTechnology Conference, pp. 900–905, May 2008.[Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=4547164
[2] J. Vasconcelos, “Survey of regulatory and technologicaldevelopments concerning smart metering in the EuropeanUnion electricity market,” 2008. [Online]. Available:http://cadmus.eui.eu/handle/1814/9267
[3] R. Olsen, “Technical considerations for broadbandpowerline (BPL) communication,” in EMC ZurichSymposium, 2005, pp. 1–6. [Online]. Available: http://preview.pserc.wisc.edu/documents/publications/papers/2005 general publications/olsen bpl paper feb2005.pdf
[4] S. Keemink and B. Roos, “Security analysisof Dutch smart metering systems,” Universityof Amsterdam, Amsterdam, 2007. [Online]. Avail-able: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.159.6314&rep=rep1&type=pdf
[5] IEEE Communication Society, “802.15.4 IEEE standardfor Information Technology-Part 15.4: Wireless MediumAccess Control (MAC) and Physical Layer Specificationsfor Low Rate Wireless Personal Area Networks (LR-WPANS),” 2003.
[6] Z. B. Alliance, “Zigbee specification,” ZigBee document053474r06, version, vol. 1, 2005.
[7] B. Lichtensteiger, B. Bjelajac, and C. Müller, “RFMesh Systems for Smart Metering: System Architectureand Performance,” Smart Grid, pp. 379–384, 2010.[Online]. Available: http://ieeexplore.ieee.org/xpls/absall.jsp?arnumber=5622071
[8] A. Liotta and G. Exarchakos, Networks for PervasiveServices: Six Recipes to Upgrade the Internet. Springer,2011, vol. 92.
[9] J. Hui and D. Culler, “Extending IP to low-power, wirelesspersonal area networks,” IEEE Internet Computing, pp.37–45, 2008. [Online]. Available: http://www.computer.org/portal/web/csdl/doi/10.1109/MIC.2008.79
[10] L. Hardy and M. Gafen, “A new highly-synchronizedwireless mesh network model in use by the ElectricCompany to switch to automatic meter reading:Case study,” 2008 5th International Conferenceon Networked Sensing Systems, pp. 31–34, Jun.2008. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=4610892
[11] A. Krohn, M. Beigl, C. Decker, and T. Riedel,“Syncob: Collaborative Time Synchronization in WirelessSensor Networks,” 2007 Fourth International Conferenceon Networked Sensing Systems, pp. 283–290, Jun.2007. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=4297432
[12] M. Ringwald and K. Romer, “BitMAC: a deterministic,collision-free, and robust MAC protocol for sensornetworks,” Proceeedings of the Second EuropeanWorkshop on Wireless Sensor Networks, 2005., vol. 00,pp. 57–69, 2005. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=1461999
[13] KEMA Consulting, “Smart Meter Requirements - DutchSmart Meter specfication and tender dossier,” 2008.
[14] D. Willkomm, S. Valentin, and T. Parker, “SimulatingWireless and Mobile Networks in OMNeT++ TheMiXiM Vision,” Proceedings of the First InternationalICST Conference on Simulation Tools and Techniques forCommunications, Networks and Systems, 2008. [Online].Available: http://eudl.eu/?id=3031
[15] D. Geelen, “Analysis and performance assessment of arouting protocol for deployment in a (meshed) smart-metering network,” 2011.
348
Figure 2: Routes generated by the simulation. 0
20
40
60
80
100
120
1e-09 1e-08 1e-07 1e-06 1e-05 0.0001 0.001
Succ
ess
rate
(per
cent
age)
Bit error rate (P[bit-error])
Measurements deliveredMeasurements acknowledged
99
99.2
99.4
99.6
99.8
100
1e-05 0.0001
Figure 3: Delivery rates.
0
20
40
60
80
100
1e-09 1e-08 1e-07 1e-06 1e-05 0.0001 0.001 0
1
2
3
4
5
Dis
tanc
e (m
eter
s)
Nr
of n
odes
(cou
nt)
Bit error rate (P[bit-error])
Nr. of hopsPath length
Transmission distance
Figure 4: Distance measures.
0
0.02
0.04
0.06
0.08
0.1
1e-09 1e-08 1e-07 1e-06 1e-05 0.0001 0.001
Dut
ycyc
le (p
erce
ntag
e)
Bit error rate (P[bit-error])
ConcentratorOther nodes
Figure 5: Duty cycles.
0
20
40
60
80
100
120
10 20 30 40 50 60 70 80 90 100
Succ
ess
rate
(per
cent
age)
Nr. of active nodes (percentage)
Energy measurementsReverse-energy measurements
Gas measurementsWater measurements
90
92
94
96
98
100
50 55 60 65 70
Figure 6: Delivery rates of different measurements.
0
20
40
60
80
100
10 20 30 40 50 60 70 80 90 100 0
1
2
3
4
5
Dis
tanc
e (m
eter
s)
Nr
of n
odes
(cou
nt)
Nr. of active nodes (percentage)
Nr. of hopsPath length
Transmission distance
Figure 7: Distance measures.
349