On the Feasibility of Satellite M2M Systems

Abstract

Machine-to-machine (M2M) communications have a very large potential market growth, particularly in the
low-end segment. Current satellite systems are not adequate to serve very large populations of low cost devices,
with low bandwidth requirements, and severe cost and energy constraints. A satellite system can bring unique
advantages in terms of cross-border coverage, availability, and security of the communication. However, it must
compete in cost with cellular and unlicensed devices, which are rapidly evolving. The high cost of the space
segment can be compensated by the high scalability of the system if the terminal cost can be kept sufficiently
low. Moreover, the low bandwidth requirements of M2M systems make reusing current infrastructure possible. In
this paper we analyze the feasibility of such M2M satellite system. We define a satellite architecture and multiple
access technique appropriate for low-cost, energy-constrained devices, and evaluate its performance in terms of
system capacity and energy usage.

Full text

1
On the Feasibility of Satellite M2M Systems
Giuseppe Cocco and Christian Ibars
Centre Tecnol`
ogic de Telecomunicacions de Catalunya – CTTC, Castelldefels – Spain
Abstract
Machine-to-machine (M2M) communications have a very large potential market growth, particularly in the
low-end segment. Current satellite systems are not adequate to serve very large populations of low cost devices,
with low bandwidth requirements, and severe cost and energy constraints. A satellite system can bring unique
advantages in terms of cross-border coverage, availability, and security of the communication. However, it must
compete in cost with cellular and unlicensed devices, which are rapidly evolving. The high cost of the space
segment can be compensated by the high scalability of the system if the terminal cost can be kept sufficiently
low. Moreover, the low bandwidth requirements of M2M systems make reusing current infrastructure possible. In
this paper we analyze the feasibility of such M2M satellite system. We define a satellite architecture and multiple
access technique appropriate for low-cost, energy-constrained devices, and evaluate its performance in terms of
system capacity and energy usage.
I. INTRODUCTION
The satellite Machine-to-Machine (M2M) communications market has a very large growth potential
[1] . Several operators are offering M2M services, and the market is rapidly developing for competing
terrestrial systems. ETSI has taken action to define requirements and a reference system architecture [2],
[3] . In the cellular arena, in 3GPP there are several active study items related to so-called machine-type
communications. In addition, in the area of short range devices, mesh networking capabilities have been
developed for the Zigbee/802.15.4 standard, aiming at large scale deployment of sensors.
Satellite systems have unique advantages for M2M communications, as well as unique challenges. They
can provide global coverage from just a few satellites, which can make the system cost effective (provided
that deployed technology is reused) with very few initial users. Moreover, satellite transmission is safer,
harder to disrupt, and easier to deploy than its terrestrial counterparts. Coverage extends to sparsely
populated areas. There is no need for extensive hardware upgrades, as in the cellular option, or for a high
density of devices, as is the case for Zigbee networks. Moreover, satellite systems can be easily integrated
with short range devices that may provide capillary network extensions from a single satellite device. One
of the main challenges of satellite M2M systems, intended for low-power devices, is providing sufficient
coverage with a modest energy budget on the device side. Large propagation distances and shadow path
loss may drain the device battery. In this respect, a low earth orbit (LEO) satellite configuration presents
clear advantages with respect to a geosynchronous orbit (GEO) option.
In this paper we propose a system architecture for a satellite M2M system targeting low end devices. We
evaluate two options: LEO and GEO, and from the link budget we derive the system parameters. We also
propose a random access scheme capable of recovering packet collisions, and evaluate its performance in
terms of throughput and energy per successful packet transmission.
II. USE CASES AND REQUIREMENTS
The M2M market encompasses a very large number of applications and is characterized by a high
fragmentation. In order to derive a set of requirements for the communications network it is necessary to
focus on a number of representative use cases. Several classifications of use cases exist. In the EXALTED
project funded by the European Commission, the main use cases are classified under the Intelligent
Ph.D. Candidate, Access Technologies, CTTC.
Area Coordinator, Access Technologies, CTTC.

2
Transport Systems area, the Smart Metering and Monitoring area, and the E-health area [4] . In [1]
, applications are classified into nine main areas (buildings, energy, consumer, health care, industrial,
transportation, retail, security/public safety, and IT/Networks). Figure 1 shows a hierarchical classification
of potential satellite M2M services based on the above.
Smart cities and infrastructure
E-health
ITS
Security and monitoring
M2M application areas
energy
Smart
roads
Fleet
manage
ment
Vehicle
diagnostics
Building
management
Advanced
agriculture
Infrastructure
monitoring
Rail
systems
logistics
Law
enforcement
Border
security
Disaster
prevention/
relief
Diagnostics/
prevention
Fig. 1. A classification of use cases for Satellite M2M.
While the requirements of M2M applications are very heterogeneous in nature, applications requiring
high throughput and low latency communications can be supported by personal communications systems
designed to support voice or Internet connectivity. On the other hand, such systems are typically too
expensive for low end applications requiring low speed, delay-tolerant data transmission. In this paper we
focus on such applications, characterized by the requirements set forth in Table I.
TABLE I
BASIC SYSTEM REQUIREMENTS.
Data rate up to 100 kbps
Activity factor Very low, below 1%
Latency Delay tolerant applications
Terminal cost Very low
Service cost Very low
Device lifetime on battery power very high (5 to 10 years)
Scalability Very high (up to billions of connected devices)
III. SYSTEM MODEL
The design objective of the proposed satellite M2M system is to provide ubiquitous coverage to
a very large population of devices that are severely cost- and energy-constrained. We propose two
different architectural approaches: a Low Earth Orbit (LEO) satellite constellation, and a single, multi-beam
Geostationary Earth Orbit (GEO) satellite. In order to support delay-tolerant traffic, a small constellation
of satellites is sufficient to ensure discontinuous coverage. The advantages of the LEO configuration are i)
a more favorable link budget, and ii) a higher elevation angle, minimizing shadowing losses. On the other
hand, a GEO satellite provides continuous coverage, which reduces latency to that incurred by propagation
and the communication protocol.

3
It is assumed that the proposed satellite M2M system will be largely self-organized, therefore a low
frequency band (L or S bands) is selected for its lower propagation loss with respect to higher bands,
which makes the system more robust against pointing errors and/or low antenna gains.
The system may also include traffic concentrators, or gateways, that provide local coverage to clusters
of sensors. This approach, typically used in terrestrial networks, allows several devices to be connected
through a local area interface to the gateway and a dual radio interface device which is also connected to
the satellite. The basic system architecture is depicted in Figure 2. A standard hardware configuration is
assumed for M2M devices, consisting of a sensor/actuator module, a processing module (microprocessor
and memory), a communications module, and a power module. Gateways must include a local area
interface as well in order to aggregate local traffic.
Fig. 2. Proposed Satellite M2M system architecture.
A. Link Budget
In this section we provide a link budget calculation in order to derive feasible system parameters,
namely transmit power, data rate, and datagram size. We focus on the uplink, which we assume is the
limiting communication segment due to the limited power and energy availability of the M2M devices.
1) LEO configuration: The effective isotropic radiated power (EIRP) at the transmitter is defined as
[5]:
EI RP ,PTGTmax
LTLF tx
,(1)
where PTis the power at the output of the transmitter’s amplifier expressed in Watts, GTmax is the
maximum gain of the transmitter’s antenna, LTis the loss due to the pointing error at the transmitter’s
antenna, while LF tx is the loss due to the antenna feeder.
The power received at the satellite is:
PR=PTGTmax
LTLF tx
1
LF S LA
GRmax
LRLF rx
,(2)
where:
LF S ,4πR
λ2
is the free space propagation loss, λbeing the carrier wavelength (λ=c/f, where c= 3 ·108m/s is the
speed of light and fis the carrier frequency) and Rthe distance between the transmitter and the receiver,
LAis the loss due to atmospheric effects (e.g.: absorption by atmospheric gas) while LRand LF rx are the
loss due to antenna misalignment at the receiver and the feeder loss at the receiver antenna, respectively.

4
Finally, the carrier-to-noise ratio is given by the following formula:
C
N0
=EI RP
LG
T1
k,(3)
where L=LF S LA,k= 1.379 ·10−3W/HzK is the Boltzmann’s constant, G=GRmax /LRLF rx is
the composite gain at the receiver’s antenna and Tis the receiver’s noise temperature. The ratio G
T
is referred to as the figure of merit of the receiver, which takes into account losses due to propagation,
satellite antenna misalignment and feeder loss at the satellite. We assume the following values for the
above parameters:
TABLE II
LINK BUDGET PARAMETERS FOR THE UPLINK.
PT20 dBm
GTmax 2 dB
f1.623 GHz
R876 km
LT1 dB
LF tx 0.5
LA0.4 dB
LR3 dB
LF rx 0.5 dB
(G/T )3 dB/K
D400 km
where Dis the satellite spot beam diameter. LRand Rhave been chosen assuming that the receiver
is locate at the edge of the spot beam and for a satellite altitude above the ground of about 780 km. By
plugging the values of table II into Eqn. (3) we find:
C
N0
= 62.7dB.
Assuming a pulse amplitude modulation (PAM) is used and using Eqn. (3), it is possible to derive the
expression for the ratio Es/N0between the energy-per-symbol and the noise floor at the satellite receiver:
Es
N0
=C
N0
Ts=C
N0
1
Rs
=EI RP
LG
T1
k
1
Rs
,(4)
where Tsis the symbol duration and Rsis the baud rate. Assuming a baud rate of Rs= 100 ksps we
find:
Es
N0


LEO
= 12.7dB.(5)
Assuming BPSK modulation and a coder ate of 1/2we have the following ratio between the energy per
bit Eband the noise floor at the satellite:
Eb
N0


LEO
= 15.7dB.(6)
2) GEO configuration: The link budget for the case of geostationary satellite can be found from Eqn.
(3) and Eqn. (4) taking into account the increase in free space propagation loss LF S due to the higher
distance. The free space propagation loss for the GEO case at 1.623 GHz is:
LF S |LEO = 188.7dB.

5
Plugging LF S into Eqn. (4) and keeping the rest of the parameters as in Table II we find:
Es
N0


GEO
=−9.7dB.(7)
As before, assuming BPSK modulation and a code rate of 1/2we have the following ratio between the
energy per bit Eband the noise floor at the satellite:
Eb
N0


GEO
=−6.7dB.(8)
B. System Parameters
In order to verify the achievability of the system requirements given in Table I, we verify the compliance,
in terms of SNR, with our proposed multiple access protocol, NCDP, described in the next section. In
[6] the FER curves are provided for a code rate 1/2 and a datagram size of about 200 bytes. In order for
the NCDP protocol to work properly a maximum decodable collision size equal to 8 may be sufficient,
which leads to a required Eb/N0of about 12 dB. According to the link budget calculations and set
of requirements set forth in the previous section, we conclude that, with little adjustments in terms of
data rate, datagram size and transmitted power, the LEO configuration may support the following system
parameters:
•Datagram size: 10-100 bytes
•Data rate: 10-100 kbps
•Transmit power: 20 dBm
As for the GEO configuration, it may not be a valid choice due to the low SNR, unless the target system
cost is significantly revised.
IV. MULTIPLE ACCESS TECHNIQUES FOR M2M
The very large amount of devices to be serviced is one of the main challengers of a satellite M2M
system. First, the amount of signaling must be low enough so that it does not represent significant overhead.
Second, a multiple access scheme that is able to schedule a very large number of small packets must
be in place. We distinguish two types of M2M traffic: schedulable and non-schedulable. Schedulable
traffic follows predetermined patterns that allow the system to reserve a large number of slots at once.
On the other hand, non-schedulable traffic is represented by random arrivals and may not be scheduled
in advance. An example of schedulable traffic are utility meter readings. If a meter must provide one
reading daily, all such readings may be scheduled well in advance. An example of non-schedulable traffic
is traffic originating from monitoring devices in the event of a failure of a system.
Schedulable traffic may be managed through a reservation mechanism, which could be initiated either
by the device or by the application manager. In such a case, the satellite transmits reservation schedules
on beacon frames and the devices wake up when their transmission window arrives.
Non-schedulable traffic is better served by a random access mechanisms. The advantage of a random
access mechanism w.r.t. scheduling transmission is that a smaller amount of overhead is required. If a
transmission is to be scheduled, each M2M device must transmit a reservation request through a random
access channel so that it can be allocated a transmission slot by the satellite gateway. However, for many
applications such a reservation packet may be of similar size than the data packet itself, therefore a
direct transmission via random access becomes more efficient. Moreover, packet acknowledgments may
be transmitted in a batch, further reducing signaling bandwidth. Scheduled traffic and random access
frames are transmitted consecutively, and each random access frame is in turn split into NCDP frames,
as shown in Figure 3

6
Downlink frame
NCDP
Frame 1
NCDP
frame 2
NCDP
frame N
Schedulable traffic frame Random access frame
NCDP
Frame 1
Fig. 3. Proposed Satellite M2M frame structure.
A. Random Access based on NCDP
A protocol for random access called Network-Coded Diversity Protocol (NCDP) was presented in [7]
. The protocol is conceptually similar to the Contention Resolution Diversity Slotted ALOHA (CRDSA)
proposed in [8] . In both schemes more than one replica of each packet are transmitted by a node in a
given frame and then a collision recovery mechanism is applied to the received frame. While CRDSA uses
a successive interference cancellation (SIC) mechanism, NCDP applies multi-user physical layer network
coding (MU-PHY NC) [9] over an extended Galois Field (GF). In the following we describe the NCDP
at the transmitter (terminal node) and at the receiver side (satellite gateway).
1) NCDP: Transmitter Side: Assume that node ihas a message uito deliver to receiver Rduring
frame f. We call active terminals the nodes that have packets to transmit in a given frame. Each message
is transmitted more than once within a frame, i.e., several replicas of the same message are transmitted.
We will give details about the number of replicas transmitted within a frame in next section. Before each
transmission, node ipre-encodes uias depicted in Fig. 4. The pre-coding process works as follows. uiis
n bits n bits
Channel
Coding
Modulation
n bits
Fig. 4. NCDP pre-encoding, channel coding and modulation scheme at the transmitter side. The message to be transmitted is divided into
sub-blocks. Each sub-block is multiplied by a coefficient αij ∈GF (2n). Coefficients αij ,j∈ {1, . . . , S}are chosen at random in each
time slot. After the multiplication, the message is channel-encoded, a header is attached and the modulation takes place.
divided into L=K
nblocks of nbits each. At each transmission a different coefficient αij ,j∈ {1,...,S},
is drawn randomly according to a uniform distribution in GF (2n). If αij = 0, terminal Tidoes not
transmit in slot j. Each of the Lblocks ur
i,r∈ {1,...,L}, is interpreted as an element in GF (2n)and
multiplied by αij . We call u′
ij the message uiafter the multiplication by αij .u′
ij is then channel encoded,
generating the codeword xij =C(u′
ij ). After channel coding, a header piis added to xij . Such header
is chosen within a set of orthogonal codeword (e.g. Walsh-Hadamard). The same header piis used for
all transmissions of node iwithin frame f, i.e., it does not change within a frame. Once the header is
attached, xij is BPSK modulated and transmitted.
The choice of the coefficients and of the header is done as follows. Node idraws a random number µ.
µis used to feed a pseudo-random number generator, which is the same for all terminals and is known at
R. The first Soutputs of the generator are used as coefficients. The header is uniquely determined by µ,
i.e, there is a one-to-one correspondence between the set of values that can be assumed by µand the set
of available orthogonal headers. The orthogonality of the preambles allows the receiver to know which
of the active terminals in frame fis transmitting in each time slot. Moreover, as the header univocally
determines µand thus the set of coefficients used by each node, Ris able to know which coefficient

7
Decoder
Linear equation in
Received
frame
Fig. 5. For each of the slots the receiver uses the
orthogonal preambles to determine the which node is
transmitting. With the same preamble the channel from
each of the transmitters in the slot to Ris estimated. The
channel hij, j ∈ {1,...,S}changes at each slot due to
phase noise, according to the channel model described in
[6] . Once the channel has been estimated, the decoder
applies MU PHY NC to calculate the bitwise XOR of
transmitted messages. The bitwise XOR corresponds to a
linear equation in GF (2n)with coefficients αij which are
known to the receiver through the header. In the figure only
bursts with non-zero coefficients are shown.
Fig. 6. The receiver tries to channel-decode all of the
occupied slots, thus obtaining a system of equations in
GF (2n). At this point, if the matrix Aof coefficients is full
rank, Rcan obtain all the original messages. If Ais not full
rank, Rcan decode the “clean” bursts (i.e., the bursts that
did not experience collision), then subtract them from the
slots where their replicas are. The procedure goes on until
there are no more clean bursts. In the figure, Trepresents
the transpose operator.
is used by each transmitter in each slot. As we we will see in Section IV-A2, this is of fundamental
importance for the decoding process. As said before, the set of headers is a set of orthogonal words,
such as those usually adopted in CDMA. The fundamental difference with respect to a CDMA system is
that in such system the orthogonality of the codes is used to orthogonalize the channels and expand the
spectrum, while in NCDP the orthogonality of the preamble is used only for determining the identity of
the transmitting node, which is obtained without any spectral expansion, as the baud rate 1/Tsis equal
to the chip rate (i.e., the rate at which the modulated symbols are transmitted over the channel) [8] .
2) NCDP: Receiver Side: The decoding scheme at the receiver side is illustrated with an example in
Fig. 5 and Fig. 6. In the example, a frame with S= 4 slots and Ntx = 3 active terminals are considered. In
each slot the receiver uses the orthogonal preamble of each burst to determine which node is transmitting
and which coefficient has been used for that burst. As described in Section IV-A1, the coefficients used
by a node in each burst are univocally determined by the preamble. The preamble can be determined at
Rusing a bank of correlators which calculates in parallel the correlation of the received signal with each
element in the set of available preambles. The preamble is also used by Rto estimate the channel for
each of the transmitters. The details about the channel estimation can be found in [6] . Once the channel
has been estimated, the decoder applies PHY NC to calculate the bitwise XOR of transmitted messages,
as detailed in [6] . The receiver tries to channel-decode the received signals using PHY NC. According
to Finite Field arithmetics, the bitwise XOR can be interpreted as a sum in GF (2n). Thus the slots that
have been correctly decoded are interpreted as a system of equations in GF (2n)with coefficients αij,
which are known to the receiver through the headers (see Fig. 5). At this point, if the coefficient matrix
Ahas full rank, Rcan recover all the original messages using common matrix manipulation techniques
in GF (2n)(see Fig. 6). If Ais not full rank, not all the transmitted packets can be recovered. However,
a part of them can still be retrieved using matrix manipulation techniques such as Gaussian elimination.
In case of rank deficient coefficient matrix an iterative cancelation process is applied over the extended
Galois Field used, as described in [7] .
Under the assumption that M2M traffic is delay-tolerant, we propose a reduced LEO constellation (down
to a single satellite) providing discontinuous coverage in order to reduce CAPEX. The main advantages
of this approach are its more favorable link budget and shadowing diversity due to satellite movement.
In addition, this scheme allows a scalable deployment, where more satellites may be added as the traffic
volume increases.

8
The GEO architecture consists of a single, multibeam GEO satellite. It can provide permanent coverage
and therefore reduce the delay of the proposed LEO configuration. However, the link budget is more
restrictive, and shadowed areas are expected increase with respect to the LEO approach.
3) Practical Aspects: Timing alignment of the received packets is required for NCDP (although certain
asynchronism ∆Tscan still be tolerated by the system, as shown in [6]). A simple method to achieve
it is to introduce a calibration phase when a devices tries to access the network for the first time. We
will assume in the following that nodes are fixed. Mobile devices may need to periodically repeat the
calibration phase. Calibration works as follows. The satellite periodically beacons a signal in which a
frame number identifier (FNid) is transmitted. The FNid is increased by one unit at each transmission;
a new FNid is transmitted each Tfseconds, Tfbeing the frame duration. We call local arrival time for
frame f(LATf) the time at which the beacon is observed at the receiving terminal. By tracking the
difference between the LATf’s and knowing the satellite height above the ground, the terminal can derive
its own distance with respect to the center of the beam in correspondence of each FNid, and thus the
difference between the own propagation delay and that of a node at the center of the beam. Note that
each terminal observes only a subset of the FNid’s beaconed by the satellite due to the spacecraft motion
and eventual deep fading.
Once the calibration process has finished the terminals enter a sleep mode, during which they collect
data but do not transmit. Terminals are woken up by a satellite beacon. A node wakes up only if the
received beacon power is higher than a given threshold. The threshold is chosen so that only nodes with
a sufficiently good channel participate to the communication (i.e., nodes in deep fade are excluded), in
order to meet the SNR requirements needed for NCDP. This beacon carries the FNid of the frame that is
going to begin. The FNid for a given frame is transmitted by the satellite always from the same orbital
position, so that once a node knows the FNid, can derive its relative delay with respect to the center of the
beam for that FNid. Once the second beam is received each node starts a timer that aims at compensating
the delay difference and having the signals of all the transmitting nodes in that frame superposing with
symbol synchronicity (within a tolerance ∆Ts) at the satellite antenna. The moment in which a node’s
timer expires is taken as the starting time of the frame for that node.
V. NUMERICAL RESULTS
In this section we present the numerical results. Our performance metrics are the normalized throughput
Φdefined as:
Φ = G(1 −Υ),(9)
where Υ∈[0,1] is the average packet loss rate (i.e, the ratio of the number of lost packets to the
total number of packets that arrive at the transmitters), and the average energy consumption per received
message η, defined as the average number of transmissions needed for a message to be correctly received
by R. We consider two benchmarks. The first one is a system that implements the contention resolution
diversity slotted ALOHA (CRDSA) protocol, which has been proposed in [8] . In CRDSA a node transmits
two or more copies of a burst (twin bursts) in different slots randomly chosen within a frame. Each of
the twin bursts contains information about the position of the other twin burst in the frame. If one of the
twin bursts does not experience a collision (i.e, it is clean) and can be correctly decoded, the position of
the other twin burst is known. These bursts may or may not experience a collision with other bursts. If it
happens, these are removed through interference cancelation using the decoded bursts. In order to do this
Rmemorizes the whole frame, decodes the clean bursts, reconstructs the modulated signals and, once the
effect of each user’s channel has been included in the reconstruction, they are subtracted from the slots
in which their replicas are located. The IC process is iterated for a number Niter of times, at each time
decoding the bursts that appear to be “clean” after the previous IC iteration. The second benchmark is a
slotted ALOHA system.
We consider an automatic repeat request (ARQ) scheme, in which a node receives an acknowledgement
(ACK) or a negative acknowledgement (NACK) from the receiver in case a message is or is not correctly

9
received, respectively. An alternative to the NACK is to having the transmitters using a counter for
each transmitted packet, indicating the time elapsed since it has been transmitted. If the timer exceeds a
threshold value (which depends on the system’s RTT), the message is declared to be lost. A node that
receives a NACK (or whose timer exceeds the threshold vale) enters a backlog state. Backlogged nodes
retransmit the message for which they received the NACK in another frame, uniformly chosen at random
among the next Bframes. We call Bthe maximum backlog time. The process goes on until the message is
acknowledged [10] . In both setups we assume a very large population of users. Furthermore, we assume
that the average SNR is high enough so that the FER at the receiver is negligible.
We evaluate jointly the spectral efficiency (average number of messages successfully received per slot)
and the energy consumption (average number of transmissions needed for a message to be correctly
received) of the schemes under study. In Fig. 7, Φis plotted against Gfor a frame size S= 150 slots
and a maximum backlog time B= 50 frames. The figure shows how Φincreases linearly with Gup to a
threshold load value. Such threshold increases with the (average) number of repetitions of the considered
scheme. The Φcurve of NCDP upperbounds that of CRDSA. The reason for this lies in the way the
decoding process is carried out by the receiver Rin NCDP. Rfirst tries to decode the whole frame, which
is feasible if the coefficient matrix Ahas rank Ntx. If the whole frame can not be decoded, then Rapplies
Gaussian elimination on A, in order to recover as many messages as possible. It can be easily verified
that Gaussian elimination in NCDP is the equivalent, in a finite field, of the IC process of CRDSA, which
is applied in the analog domain. In order to compare jointly the spectral and the energy efficiency of the
0.2 0.4 0.6 0.8 1
0
0.2
0.4
0.6
0.8
1
G
Φ
NCDP d= 2
NCDP d= 3
NCDP E[d]=6.795 (p= 0.0453)
CRDSA d= 2
CRDSA d= 3
Slotted ALOHA
Fig. 7. Normalized throughput Φvs normalized traffic load Gin a system with retransmission. In the simulation the frame size was set
to S= 150 slots while the maximum backlog time was set to B= 50 frames.
different schemes, we plot the curves for the normalized throughput vs the average energy consumption
per received message η, which is shown in Fig. 8. The increase in throughput coming from an increased
number of transmissions implies a higher energy consumption for a given transmitter in a given frame.
However, this does not necessarily imply a loss in energy efficiency. As a matter of facts, the simulation
results we are going to present show that there is not a scheme that outperforms the others in terms of
both energy and spectral efficiency, but which scheme is best depends on the maximum throughput we
want to achieve. In Fig. 8 we see that SA achieves a higher throughput with a lower energy consumption
with respect to the other schemes in the region Φ<0.35. In the region Φ>0.35, instead, both NCDP
and CRDSA achieve a higher throughput with lower energy consumption with respect to SA. NCDP and
CRDSA behave almost in the same way in the case of 2repetitions, achieving a maximum throughput
of 0.5for an average energy consumption of 2. In the case of 3repetitions NCDP achieves a maximum
Φof 0.7, higher than CRDSA, for which the peak value is 0.6, for η= 3. In the NCDP scheme with a
retransmission probability of p= 0.0453 a peak throughput of 0.8is achieved in correspondence of an
average energy consumption of η= 6.795. For comparison, we also show the throughput-energy curve

10
100101102103104105
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
η
Φ
NCDP d= 2
NCDP d= 3
NCDP E[d]=6.795 (p= 0.0453)
NCDP E[d] = 149.415 (p= 0.9961)
CRDSA d= 2
CRDSA d= 3
Slotted ALOHA
Fig. 8. Normalized throughput vs average energy consumption per decoded message for S= 150 and B= 50 frames.
for NCDP in case of p= 0.9961, i.e., coefficients αij are chosen uniformly in GF (28). The high pleads
to a high throughput, but also to a high energy consumption, with a minimum of η= 149.415. Moreover,
we note that the gain with respect to the scheme with p= 0.0453 is negligible (about 5%), especially
when compared to the energy saving of about 95% of this last one.
Comparing Fig. 7 and Fig. 8 it is interesting to note how the energy consumption does not change with
the load up to a threshold value of G, below which the normalized throughput increases linearly, i.e., there
are almost no losses and no retransmissions. The total offered load Gincreases with the geographical
density of the nodes and with their activity factor, which are parameters that, for certain applications, can
be either known a priori or obtained from the network after its deployment. Knowing the operating load,
the number dof repetitions can be set so as to minimize the energy cost per transmitted message. For
instance, according to the figures above, if G= 0.7it is better to use d= 2 repetitions instead of 3, while
in case G= 0.8d= 3 repetitions would be the best choice. dcan be reset easily by means of a new
calibration phase in case the node density changes over time. In case the sensors are battery-operated this
elements have a fundamental impact in the lifetime of the system and must be taken into account.
VI. CONCLUSIONS
In this paper we outlined an architecture for a M2M satellite system targeting low end applications.
The system relies on a small LEO constellation and a newly proposed collision resolution method for
random multiple access. According to the considered link budget obtained using realistic parameters we
showed that the proposed M2M system requirements can be met in a LEO satellite configuration, while
the GEO configuration requires a significant relaxation of the cost/complexity constraints on the nodes.
Our future work will focus on the diversification of the random access mechanism according to the
terminal types and on an accurate dimensioning of the system parameters in order to increase the link
margin for fading and implementation losses.
ACKNOWLEDGEMENT
The authors would like to thank Nader S. Alagha of the European Space Agency for the fruitful
discussions and supervision during the redaction of the paper.
This is the pre-peer reviewed version of the following article: “G. Cocco, C. Ibars, On the Fea-
sibility of Satellite M2M Systems, in proceedings of the 30th AIAA International Communications
Satellite Systems Conference (ICSSC 2012), 24-27 September 2012, Ottawa (Canada)”, available at
http://arc.aiaa.org/doi/abs/10.2514/6.2012-15074.

11
REFERENCES
[1] “Satellite machine-to-machine services market study,” European Space Agency Contract Report, 2011.
[2] “Machine to machine communications (m2m); m2m service requirements,” ETSI TS 102 689.
[3] “Machine to machine communications (m2m); m2m functional architecture,” ETSI TS 102 690.
[4] “Business models, use cases & technical requirements,” Deliverable D2.1, EXALTED Project, www.ict-exalted.eu, 2011.
[5] G. Maral and M. Bousquet, Satellite Communications Systems - systems, techniques and technologies, John Wiley & Sons Ltd, 2009.
[6] G. Cocco, N. Alagha, C. Ibars, and S. Cioni, “Practical issues in multi-user physical layer network coding,” in accepted at Advanced
Satellite Mobile Systems Conf., Baiona, Spain, Sep. 2012.
[7] G. Cocco, N. Alagha, C. Ibars, and S. Cioni, “A network-coded diversity protocol for collision recovery in slotted aloha networks,”
submitted to IEEE Journal on Selected Areas in Comm.
[8] E. Casini, R. De Gaudenzi, and O. d. R. Herrero, “Contention resolution diversity slotted ALOHA (CRDSA): An enhanced random
access scheme for satellite access packet networks,” IEEE Trans. on Wireless Comm., vol. 6, no. 4, pp. 1408–1419, Apr. 2007.
[9] G. Cocco, C. Ibars, D. G¨und¨uz, and O. d. R. Herrero, “Collision resolution in slotted ALOHA with multi-user physical-layer network
coding,” in IEEE Vehicular Technology Conf. (VTC Spring), Budapest, Hungary, May 2011.
[10] L. Kleinrock and S. Lam, “Packet switching in a multiaccess broadcast channel: Performance evaluation,” IEEE Trans. on Comm.,
vol. 23, no. 4, pp. 410–423, Apr 1975.