-
992 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
Multihop R-ALOHA for IntervehicleCommunications at Millimeter
Waves
Roberto Verdone, Member, IEEE
AbstractWith reference to road transport information
(RTI)applications, such as cooperative driving, short-range
intervehiclecommunications in a highway environment are
investigated inthis paper. The research in this field indicates the
suitability ofthe 6064-GHz band.Due to the distributed nature of
the intervehicle communication
system, an R-ALOHA protocol is considered; multihop (MH)
andsingle-hop (SH) strategies are compared. Network performanceis
assessed by considering the joint impact of random
access,interference, thermal noise, propagation, and packet
captureeffect. Several figures of merit are analyzed and discussed:
packetsuccess probability (PSP), system stabilization time (SST),
firstsuccess time (FST), and deadline failure probability (DFP).
Net-work performance is evaluated either by an analytical
approachor by a software tool able to simulate a one-lane highway
sce-nario. Both steady-state and transition situations are
considered.System performance in terms of PSP (in the presence of
two-ray Rice fading, noise, and interference with antenna
diversityand selection combining) is analytically evaluated to
validate thesimulation tool and to prove the suitability of an MH
networkstrategy. The simulation approach allows the evaluation of
theimpact of protocol parameters on network performance,
withreference to nonsteady-state situations.
Index TermsHighway, intervehicle, millimeter waves,
multi-hop.
I. INTRODUCTION
WITH THE purpose of designing a road traffic man-agement system,
either in urban or in highway envi-ronments, intervehicle
communications will play a relevantrole for road transport
information (RTI) applications such ascollision avoidance and
cooperative driving, aimed at improv-ing traffic safety and
efficiency [1], [2]. Several EuropeanResearch Programs (such as
DRIVE and PROMETHEUS)have dedicated efforts to this field in the
past years. Sys-tem designers are considering higher and higher
frequencyranges having larger bandwidth capacity; more
precisely,microwave and millimeter wave systems are investigated
inEurope [3][5].In a highway environment, information exchange
(about
speed, acceleration, etc.) between vehicles has to be
allowedeither in the presence or in the absence of fixed
infrastructure:this can be achieved by means of short-range
intervehiclecommunication links [6], [7]. In this sense, the use of
the
Manuscript received April 7, 1996; revised September 9, 1996.
This workwas performed under contract with Progetto Finalizzato
Trasporti 2, CNR,Rome, Italy.The author is with CSITE-CNR,
University of Bologna, 40136 Bologna,
Italy.Publisher Item Identifier S 0018-9545(97)05116-5.
6064-GHz band is suggested by the presence of a peak inthe
oxygen absorption at those frequencies, allowing efficientspatial
filtering effects. In Europe, the 63.5-GHz band has beenrecommended
[8].In this paper, a short-range communication system at 63.5
GHz for highway environments and cooperative driving
appli-cations, is considered. Due to the need for distributed
networkmanagement, a suitable multiple-access protocol has to
beproposed. Moreover, since propagation impairments can causehigh
levels of unavailability of the service, proper
transmissiontechniques have to be exploited [9], [10].In the
literature, some recent papers have been dedicated
to intervehicle communication systems at millimeter
waves[9][14], but usually investigating transmission or
networkaspects separately.The aim of this paper is to evaluate
network performance
by taking propagation impairments, transmission
techniques,random access and network structure into account: the
effectsof fading and cochannel interference are also considered.In
[11], narrowband and spread-spectrum modulation sys-
tems were compared in the framework of an S-ALOHAprotocol, and
the former showed superior performance in thegiven scenario. In
[12] and [13], an S-ALOHA network ina real environment was
investigated. However, for coopera-tive driving applications,
vehicles have to send their statusinformation to the neighborhood
mobiles periodically; hence,a protocol allowing a form of channel
reservation shouldbe considered. An R-ALOHA network was
investigated in[15], but an ideal channel was assumed for
performanceevaluation. R-ALOHA and its modifications were
proposedfor intervehicle communications also in [6] and [7]This
paper starts from the conclusions drawn in [15], and
its first aim is to show the impact of the propagation channelon
network performance when R-ALOHA is used; a one-lanescenario, where
vehicles are separated by a fixed distance,is considered. Frame
synchronization is assumed, which canbe obtained either by means of
roadside infrastructures sup-porting the intervehicle system, or by
exploiting a distributedalgorithm [16].The number of mobile
terminals that should receive the
information referred to a given vehicle [defining the
so-calledzone of relevance (ZOR)] is a parameter of relevance, from
thepoint of view of both the application and the
communicationprotocol [17]; moreover, the information could be
transmittedeither by broadcasting to all the vehicles in the ZOR,
orby means of multihop (MH) links, where mobile terminalsretransmit
received packets. In this paper, MH and single-hop
00189545/97$10.00 1997 IEEE
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
993
(SH) R-ALOHA systems are compared, in order to underlinethe
suitability of the MH strategy in the given context.In the paper,
several figures of merit are analyzed: packet
success probability (PSP), system stabilization time (SST),
firstsuccess time (FST), deadline failure probability (DFP).
Theperformance is assessed as a joint effect of fading, noise
andinterference: suitable channel characterization at 63.5 GHz
isgiven by means of a two-ray Rice model taking the effectof the
road-reflected wave into account. With the purpose ofevaluating
network performance, two means are exploited.1) An analytical
procedure [14], [18][20] allowing the
evaluation of PSP in the joint presence of thermalnoise,
cochannel interference, Rician fading, and antennadiversity with
selection combining.
2) A simulation tool able to model the one-lane
highwayenvironment under investigation. It allows the simulationof
nonsteady-state situations.
Both the analytical and the simulation approach take intoaccount
channel impairments due to fading, antenna diversity,the
mo-demodulation techniques employed and the strategy ofreceived
power measurement; the packet capture effect, in theframework of
the R-ALOHA protocol, is also considered.As a first step in the
paper, PSP is evaluated by means
of the analytical procedure. Its determination allows
propertesting of the simulation tool results; moreover, starting
fromthe analytical approach, a suitable comparison between
theperformance of the SH and MH strategy can be carriedout. Then,
with the aim of stressing the impact of protocolparameters on
network performance in the real environmentconsidered,
nonsteady-state situations are analyzed through theuse of the
simulation tool.In conclusion, the combined approach enables proper
inves-
tigation of system sensitivity, in a real environment, to
severalparameters such as the number of vehicles in the network,the
number of slots per frame, bit rate, intervehicle distance,the SH
or MH strategies, power requirements, etc. The impactof channel
propagation, movement of vehicles and antennadiversity on network
performance is evaluated.The paper is organized as follows. In
Section II, the com-
munications scenario is introduced. In Section III, the
maincharacteristics of the intervehicle link at 63.5 GHz are
sum-marized. Section IV introduces the analytical method for
linkperformance evaluation proposed in [14]. Then, the assump-tions
regarding the network model are described (Section V).The
performance comparison between an SH and an MHapproach is shown in
Section VI. Section VII presents thesimulation tool for analyzing
network behavior in nonsteady-state situations. Finally, numerical
results are given.
II. THE COMMUNICATIONS SCENARIOA one-lane highway scenario is
investigated: a group ofmobiles separated by an intervehicle
distance is
considered, as shown in Fig. 1. From a communications pointof
view, mobile users are distinguishable by means of
theiridentification number (ID).A parameter of relevance in network
design is the number
of terminals, which have to receive the information about a
Fig. 1. One-lane highway scenario.
Fig. 2. SH and MH strategies.
given vehicle, in its neighborhood. It defines the ZOR [17].
Inthis paper, it is assumed that the status of a given mobile has
tobe known by the closest terminals: of these behindand ahead.
Different values of are considered.An R-ALOHA protocol is
investigated: the time resource is
subdivided into frames of duration seconds, each of
themconsisting of slots. Each vehicle exploits the radio channelto
transmit packets containing its status to the closestmobiles in its
ZOR, according to (see Fig. 2) the following.1) An SH strategy: the
vehicle broadcasts the packets to
the mobiles directly2) An MH strategy: the vehicle transmits the
packets to
the two closest terminals ; these retransmit theinformation in
the following frame (together with theirown data), playing as
repeaters.
For cooperative driving applications, a channel bit rate inthe
order of 10 Kb/s is sufficient to allow the necessaryinformation to
be exchanged [17]. However, data have to beexchanged cyclically, in
order to permit proper control oftraffic flow, with a period of
about 100 ms [17]. Thus, theframe duration has to be in the order
of 100 ms, whereas slotduration, will depend on In the paper, as a
referencecase, the overall bit rate is fixed at Mb/s.The protocol
has distributed management, and, as a conse-
quence, channel (slot) reuse is obtained, at distancewhere is an
integer (random) value.
III. THE INTERVEHICLE LINK AT MILLIMETER WAVESDue to the choice
of a contention protocol and to the
implicit reuse strategy introduced by the network, each packetis
received in an interfered environment, where one vehicleplays the
role of useful transmitter and are interferers. Ingeneral, only the
two closest interfering terminals need to beconsidered.
A. Vehicle EquipmentEach vehicle is equipped with a
communication device and
four directional antennas (see Fig. 3): two of them radiatingin
the backward direction and two in the forward.Each couple of
antennas represents a two-branch diversity
system, where the two elements are vertically separated by
adistance The two couples are separated by a distance
In [9], the efficiency of height in comparison withhorizontal
diversity was shown. Let us consider the link
-
994 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
Fig. 3. Vehicle equipment.
Fig. 4. Two-ray Rice model.
between a transmitting and a receiving mobile: antenna
heightswith respect to flat terrain are assumed to fluctuate dueto
ground unevenness around mean values and
for the transmitting vehicle, andfor the receiver. The
instantaneous values of
heights will be denoted by and respectively,or more generally by
and Fluctuations are modeled bya sinusoidal law with frequency 1 Hz
and peak deviationThe transmitted power is denoted by transmitter
and
receiver antenna gains by and , respectively. Thereceiver is
characterized by means of its noise figure
B. Propagation ModelAt 63.5 GHz, the specific attenuation due to
oxygen absorp-
tion is roughly equal to dB/Km. Moreover, rainfallcan introduce
large values of the specific attenuation due torain, Let us denote
by the overallspecific attenuation.In the given scenario, a flat
fading channel can be assumed
[4], [22]; constant power at each antenna is received during
aslot time. On the other hand, it is assumed that in
consecutiveslots power samples are independent [12]. In the
following,indicates the useful received power. Let us also denote
bythe th interfering received power.In [9], the channel model for
intervehicle links in the
6064-GHz band in highway scenarios, was described; a two-ray
Rice model, based on the coherent cumulation of the directand
road-reflected (with reflection coefficient paths, andthe
incoherent addition of a multipath power component, waspresented.
For the sake of completeness, the relations holdingin the two-ray
Rice model are shown [9].With reference to a generic link (either
useful or interfering,
see Fig. 4), let us denote by the distance between
thetransmitting and receiving mobiles (referred to, e.g.,
forwardantennas); the distance between antennas is
(1)
where antenna number 1 is assumed to be the transmitting
one,whereas can denote either or In the given scenario,
, where is an integer. A reference received powercan be
introduced as the free-space received power at
distance , which depends on link budget parameters andwavelength
[9].As far as the direct path is concerned, let us denote by
and the useful received power and link range,
respectively;similarly, and are the th interferer direct path
receivedpower and link range, respectively Thedistances between
antennas are denoted by and Wehave [21]:
(2)
(3)
(4)
The total received average power for the generic link isgiven by
the sum
(5)
where is the power due to the direct and road-reflectedwaves
(when present), the average multipath power. De-pending on the
presence of line-of-sight (LOS) conditions, wemay have two
different situations, described in the following.1) LOS link: in
the presence of direct and road-reflected
paths, the Rice factor describing the statistic of thereceived
power is given by
(6)
where is the road-reflection coefficient, thereflection-free
(direct) path power (corresponding to
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
995
and for the useful and th interfering signals,respectively),
and
(7)
is the phase shift between the direct and the road-reflected
waves. Under these conditions, the ratio
(8)
and are the two degrees of freedom of the model.2) NonLOS (NLOS)
link: when the direct and road-
reflected paths are obstructed, the received signalstatistic is
Rayleigh distributed; hence
(9)
whereas the average multipath (overall) power canbe determined
by choosing a proper value for the pa-rameter
(10)
where represents the received power in case of freespace
propagation (corresponding to or Underthese conditions, is the only
degree of freedom of themodel.
In both cases, the probability density function of the
short-term useful power can be generally written as
(11)
where is the average multipath received power [generallydenoted
by in (5)] from the useful user given by
(12)
where the Rice factor, which, from (6) and (8), is equal
to(13)
and represents the modified zero-order Bessel function ofthe
first kind.Similar expressions hold for the th interfering signal
by
replacing and with andIt is worth noting that the useful and
interfering signals
can generally come from a LOS or a NLOS link, dependingon the
receiver and transmitter relative positions in the lane;since in
this paper we assume a fixed (when is chosen)monodimensional
scenario, the following considerations canbe made.1) When the
signal comes from one of the two closest
vehicles, we have a LOS link, characterized by a givenRice
factor and (two-ray Rice fading) [9].
2) When the signal comes from vehicles at a larger dis-tance, we
have a NLOS link because of the obstructiondue to the closest
vehicles, so and canbe chosen (Rayleigh fading).
TABLE ISYSTEM AND LINK BUDGET PARAMETERS
EXPLOITED IN THE PAPER
Some analytical evaluations based on the experimentalresults of
[4] suggest the values dB andfor LOS conditions and dB for NLOS.As
far as the system and link budget parameters (trans-
mitted power, antenna gains, etc.) are concerned, the
valuesreported in Table I, based on existing RF devices, are
exploitedthroughout the paper.
IV. LINK PERFORMANCELink performance can be defined in terms of
PSP and outage
probability.
A. Capture ModelA threshold capture model is assumed in this
paper: having
fixed the desired link quality in terms of bit-error
probability,since received power in a packet time is assumed to
be
constant, the PSP can be described as
(14)where
conventional signal-to-noise ratio;minimum value of required to
obtainthe desired link quality in the absence ofinterference;total
interference short-term received power;signal-to-interference
protection ratio relatedto
The signal-to-noise threshold and the protection ratiodepend on
the transmission system and the multiple-accessmethod considered,
the countermeasures employed, and thequality desired. The
conventional signal-to-noise ratio at thereceiver can be defined
as
(15)
Noise power is evaluated over a bandwidth equal to the bitrate ,
starting from the two-sided power spectral densityof thermal noise
, which depends on the receiver noisefigure Let us also define the
threshold useful power
In [20], it was shown that for a minimum shift keying(MSK)
system with limiter-discriminator (LD) detection, the
-
996 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
Fig. 5. Outage probability as a function of interferer position:
(a) without and (b) with height diversity.
values dB and dB correspond to a bit-error probability (when two
interferers are present)Here, we assume that this quality level is
sufficient (in thepresence of a suitable coding scheme) to capture,
with highprobability, the packet [23]. So, in the numerical
results,as a reference case, dB and dB arefixed.In [14] and [19],
an analytical procedure to derive outage
probability defined as
(16)has been proposed. Since from (14) and (16) we have
(17)the same analytical procedure of [14] and [19] can be
exploitedto derive PSP. In Appendix A, the approach proposed in[14]
and [19] is summarized; it takes thermal noise, Ricefading, and
interference into account in the presence of antennadiversity with
selection combining: the antenna having thehighest level of
measured total short-term received poweris connected to the
receiver [24]. The analytical method isapproximated: the validity
of the approximation was checkedin [14], [18], and [19] by
comparison with Monte Carlosimulation results and other
approaches.
B. Numerical EvaluationsIn order to stress the importance of the
combined effects
of the propagation channel and the multiple-access protocol,the
transmission of a packet from the closest vehicle, instatic
conditions (that is, neglecting antenna fluctuations),
isconsidered.By fixing the intervehicle distance m, in Fig. 5
outage probability is shown as a function of the position in
the lane of the closest interferer with respect to the
receiverThe relative position of the useful transmitter is equal
toone (that is, a LOS useful link is considered). The performanceis
evaluated either in the presence or in the absence ofantenna
diversity. The propagation channel model describedin Section III is
used: the values of system parameters arethose described in Table
I. Fig. 5 shows that is almostunitary when (since the useful and
the interferinglinks have the same range), that is, the PSP is
about zero;however, because of the obstruction of the direct and
road-reflected waves for interfering signals coming from
distantvehicles, is very small (that is, PSP is almost
unitary)whenNow, in order to emphasize the role played by the
direct
path in the link performance, results are shown in the absenceof
interference as an ideal case to which the network tends.Fig. 6
shows as a function of , with diversity; the choiceof the link
budget parameters allows large value of PSP upto several hundred of
meters. In Fig. 7, is displayed as afunction of the position in the
lane of the transmitter withrespect to the receiver with diversity.
Several values of
are exploited.
V. THE NETWORK MODELA brief description of the R-ALOHA protocol
considered
is given in this section.At first, a vehicle contends for the
radio channel in a
frame by transmitting in a randomly chosen slot a packetin both
directions (backward and forward) and waiting foracknowledgments
(acks) from the closest terminalsof these behind and ahead). This
is the contention phase.Upon reception of all the acknowledgments
during the frame,the vehicle will use the same slot in the
following frames; this
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
997
Fig. 6. Outage probability as a function of intervehicle
distance for a LOS link in the absence of interference.
Fig. 7. Outage probability as a function of useful transmitter
position in the absence of interference.
will be the contention-free phase. If an ack is missing, a
slotwill be randomly chosen in the next frame and the
contentionphase will continue.The choice of the slot to be used for
transmission in a frame
during the contention phase is made on the basis of a
tableindicating which slots are busy and which are free. This
tableis achieved by means of total received power measurements,and
consequent comparison with a given threshold carriedout by each
vehicle.
In the contention-free phase, a packet can be lost due tofading;
if this happens, the same slot will be exploited in thefollowing
frame until a chosen time-out expires.The following simplifying
assumptions are made throughout
the paper.1) Packets: they have the same duration of one slot.2)
Framing: perfect frame synchronization is obtained by
means of a global time reference.
-
998 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
3) Acknowledgments: if a terminal receives a packet cor-rectly,
its ack has a probability of success equal to one;acks can be
transmitted by means of a time division du-plexing (TDD) technique
and a dedicated frame section.It is not the intention of this paper
to define how acksshould be sent to transmitting vehicles.
4) Power measurements: they are carried out during
non-transmitting slots; hence, the slot just exploited can bereused
in the next frame. The power measured is the sumof the short-term
received powers at the two antennas.
5) Mobile identity: each vehicle has knowledge of theID of the
closest terminals ahead and thebehind; this can be obtained by
means, e.g., of a pe-riodical procedure in which, at regular
intervals, eachmobile transmits in a round-robin fashion a short
packetinforming the closest terminals of its location [25].Thus,
vehicles know how many acks are expected tobe received.
6) Destination addressing: each transmitted packet containsthe
ID of vehicles which have to receive it; a mobilediscards any
packet which is not intended for it orit retransmits the
information in the following frame(together with its own data) if
the MH strategy is usedand the packet is to be received by other
vehicles in itsneighborhood [25].
VI. SINGLE-HOP AND MULTIHOP STRATEGIESThe aim of this section is
the comparison between the
performance of an SH and an MH strategyin a R-ALOHA framework
and a monodimensional scenario.Let us assume a steady-state
situation, where each vehicle
in the lane is in the contention-free phase and the
distributionof the time resource is optimal: vehicle 1 transmits
during slot1, vehicle 2 during slot 2, , vehicle during slot
,vehicle during slot 1, and so on. That is, the sameslot is reused
at distance Let us also consider, asan example, a value of large
enough to have noise-limitedlinks (Fig. 5 shows that is sufficient
for this purpose,i.e., has to be chosen).The choice of mainly
affects the ability of the network
to react to situations of danger (e.g., a warning from a
vehicleahead due to an accident) and the PSP.Let us define, as a
quality measure for the comparison, the
steady-state DFP (SS-DFP) defined as- (18)
where is the packet delivery delay and is a giveninteger number.
The SS-DFP obviously depends on PSP,which, as shown in Fig. 7, is
highly affected by the relativepositions of active terminals in the
lane; for this reason,in the following PSP(m) will denote the PSP
at distance
At first, let us evaluate PSP(m) for SH and MH strategy.1) SH:
for the sake of simplicity, let us consider only
the transmission to the farthest vehicle in the ZORwe have
(19)
where the dependence of on the relative positions oftransmitter
and receiver has been emphasized.
2) MH: let us assume that a packet to be repeated is
retrans-mitted only in the subsequent frame; by concentratingon the
farthest vehicle in the ZOR, we can calculate the
as a function of the PSP(1) of each hop(20)
Now, the SS-DFP, whose dependence on and will beexplained in the
following, can be evaluated as
-
(21)
SH: it can be easily found that
(22)hence
-
(23) MH: if is smaller that we have SS-DFP since at least frames
areneeded for hops; on the other hand, when
(24)
hence
-
(25)
The strategy having the smallest value of SS-DFPis the most
appropriate since it determines the largest valuesof probability of
having a fast transmission of informationthrough the lane. As an
example, Figs. 8 and 9 show SS-DFP as a function of for
intervehicle distance equal to 200m, in the interference-free case
considered in Fig. 7, by fixingdifferent values of From the
figures, it can be noted that forvalues of slightly larger or equal
to an MH strategyoffers better performance than SH, in terms of
SS-DFP.Let us note that the delay introduced by the MH scheme
can
be larger, in mean; however, if frame length and bit rate
aresuitably chosen, the delay remains within acceptable bounds.For
this reason, in the following an MH strategy will be
considered. On the other hand, it is worth noting that theimpact
of the strategy on network performance has to beconsidered also in
nonsteady-state conditions. This will bedone in the next
sections.
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
999
Fig. 8. Steady-state deadline failure probability as a function
of N for the SH and MH strategy; div
= 200 m and Nz
= 4 and 6.
Fig. 9. Steady-state deadline failure probability as a function
of N for the SH and MH strategy; div
= 200 m and Nz
= 8 and 10.
VII. THE SIMULATION TOOLA software simulation tool has been
developed to analyze an
R-ALOHA network in a one-lane highway scenario; the aim ofthe
computer program is the evaluation of the impact of proto-col
parameters (such as, e.g., the number of slots per frame) onsystem
performance. The multiple-access protocol, the channelpropagation
and packet capture model implemented are thosedescribed in previous
sections.
Fig. 10. Test scenario for validation purposes.
For the purpose of testing and validating the computerprogram,
with particular emphasis on the channel propagation
-
1000 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
Fig. 11. PSP as a function of useful distance when interference
is at di
= 200 m; comparison between the analytical approach (line) and
thesimulation results (circles).
Fig. 12. Probability distribution of SST with reference to
number of frames; the impact of Nr
;N
s
= 16; N
v
= 9; and div
= 200 m.
and packet capture modules, the following situation has
beeninvestigated, by means of either the analytical approach
pre-sented in Section IV, or simulation: PSP is found as a
functionof useful distance for a LOS link between a transmitting,T,
and a receiving, R, mobile, when an interfering vehicle, I,is
transmitting from a fixed distance m on an NLOSlink (see Fig. 10).
Fig. 11 shows the comparison obtained
by means of the two approaches for some values of
usefuldistance. The results validate the computer program.The
simulation tool enables the evaluation of the probability
distribution of some figures of merit defined in the
following.Although it can be exploited to investigate either
steady-stateor nonsteady-state conditions, in this paper the latter
case isanalyzed since the intervehicle communication network
must
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
1001
be able to react quickly to unpredictable situations. Thus,
aworst case situation is considered, as in [15]: let us assume
thatall vehicles in the lane simultaneously compete for a
channelstarting from a memoryless state, that is, no information
onfree and collision slots is available.The probability
distributions of SST and FST are evaluated
as a function of the number of slots per frame the numberof
vehicles in the lane intervehicle distance and thenumber of mobiles
which have to receive a packet1) SST is defined as the number of
frames that elapse until
all the vehicles in the lane have successfully lockedonto a slot
(by transmitting a packet and receiving theexpected number of
acknowledgments: let us call thisthe success event) [15]; it
depends on
2) FST is defined as the number of frames that elapseuntil a
given mobile succeeds in transmitting its packet(success event)
[15]. It is not a function of as longas is sufficiently large.
Their probability distributions are estimated by counting
thenumber of times the observed success event happens forthe first
time at the th frame from the start of simulation:the probability
is given by the ratio between
and the total number of iterations,can be determined in a
similar way.From the above-mentioned figures of merit, the
nonsteady-
state DFP (NSS-DFP) can be defined as- (26)
where is the maximum delay required in the
successfultransmission of one packet.It is worth noting that SST,
FST, and NSS-DFP are defined
with reference to the number of frames, as in [15]. However,when
different network configurations are investigated byvarying the
number of slots per frame, nonhomogeneousparameters are obtained.
So, in the next section when isconsidered as a variable parameter,
SST and FST are describedwith reference to the number of slots
instead of number offrames
VIII. NUMERICAL RESULTSThe simulation campaign was based on
iterations
for each set of input parameters. In this section, the
resultsobtained are presented.Some of the system parameters have
been fixed: they
are summarized in Table I. It is worth recalling that
thepropagation model is chosen in accordance with Section
III.Height diversity with selection combining, fluctuations
ofantennas due to ground unevenness, and two-ray Rice fading(in LOS
links) are considered.Power thresholds have been fixed, and their
values are re-
ported in Table II; the value of has not been optimized,
sofurther investigation could lead to slightly better
performanceresults (see [26]).
A. The Impact ofFig. 12 shows the probability distribution of
SST (in
terms of the number of frames) when varying for
TABLE IIVALUES OF POWER THRESHOLDSEXPLOITED IN THE
SIMULATION
TABLE IIILIST OF ACRONYMS EXPLOITED
THROUGHOUT THE PAPER
and m. The impact ofon the reaction time of the network is
evident. It is clearthat by choosing an MH network , a much
fasterallotment of channels is obtained.Hence, in the following
will be fixed.
B. The Impact of Channel PropagationFig. 13 shows the
probability distribution of SST when
fixed and the value ofthe intervehicle distance varies from 20
to 400 m. InFig. 14, the same performance results are shown in the
ab-sence of antenna diversity. In both cases, the curves inthe
figures are almost indistinguishable; this underlines thefact that
up to several hundred of meters, network behavioris determined by
protocol optimization rather than by linkperformance.
C. The Impact ofIn Fig. 15, the probability distribution of FST
in terms of
the number of frames is displayed; andm are fixed. The minimum
value of the number neededto approximate an infinitely long lane of
vehicles has beenchosen as a function of the number of slots per
frame.varies from 6 up to 12. From the figure, it seems that
byincreasing , smaller and smaller values of the mean FSTcan be
obtained. However, as already noted, when varying
, the number of frames is not a homogeneous measureof time. So,
Fig. 16 shows the same results with reference tonumber of slots.
Finally, in Fig. 17, the NSS-DFP is depictedas a function of delay
expressed in time slots. This lastfigure points out that when is
larger than ten, no increasein the performance is achieved.
IX. CONCLUSIONSMH and SH R-ALOHA networks for intervehicle
communi-
cations in a highway environment at 63.5 GHz, with referenceto
cooperative driving applications, have been studied.
Severalconclusions can be carried out. Table III reports the list
ofacronyms exploited throughout the paper.
-
1002 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
Fig. 13. Probability distribution of SST with reference to
number of frames; the impact of div
;N
s
= 16; N
v
= 9; and Nr
= 2:
Fig. 14. Probability distribution of SST with reference to
number of frames; the impact of div
when antenna diversity is neglected; Ns
= 16;N
v
= 9;
and Nr
= 2:
1) R-ALOHA protocols can be exploited for the
applicationconsidered since they provide good reaction times
todangerous situations and offer an efficient means for thepurpose
of allowing periodical data exchange betweenvehicles.
2) MH strategy performs better than SH in the givenenvironment
because of the monodimensional structureof the scenario.
3) A link budget has been found, based on existing RFdevices,
giving good link performance up to severalhundred of meters.
4) The impact of channel characteristics, given that onlyLOS
links between adjacent vehicles in the lane areexploited, is
negligible.
5) The number of slots per frame can be chosen accordingto the
performance analysis carried out in Fig. 17, where
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
1003
Fig. 15. Probability distribution of FST with reference to
number of frames; the impact of Ns
;N
r
= 2 and div
= 200 m.
Fig. 16. Probability distribution of FST with reference to
number of slots; the impact of Ns
;N
r
= 2 and div
= 200 m.
it has been shown that a number of slots per framelarger than
ten does not improve system performance.It is worth noting that
different values of couldlead to different numerical conclusions;
moreover, theimpact of on system performance also depends onthe
intervehicle distance. Nevertheless, from the analysisshown, the
presence of an optimum value of the numberof slots per frame is
underlined.
On the other hand, some of the conclusions could depend onthe
given one-lane scenario; the effect of the same parameterssuch as,
e.g., on system performance could be different ina multilane
scenario. This will be investigated in future work.
APPENDIX ATHE ANALYTICAL PROCEDURE TO DERIVE PSP
This methodology is based on considering the probabilitydensity
function of total interfering short-term received
-
1004 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 46, NO. 4,
NOVEMBER 1997
Fig. 17. NSS-DFP with reference to number of slots; the impact
of Ns
;N
r
= 2 and div
= 200 m.
power to be Gaussian, with mean
(27)
and variance
(28)
From this assumption, an expression for is determined.In the
absence of antenna diversity, we have [19]
(29)
Let us now consider antenna diversity and denote by thetotal
received short-term power
(30)Moreover, in order to distinguish signals at each
antenna,
and denote the useful total interference andtotal short-term
power at the th antenna placed at height
respectively Signals are assumed to beindependently faded at
each antenna.By defining at the th antenna, an
outage condition occurs if the received useful power isless than
the threshold or the ratio between useful andtotal interference
received power is less than Letus denote by a Boolean variable,
which indicateswhether an outage condition occurs at the th
antenna.
Due to the combining method chosen, at the combineroutput outage
can occur in three different cases.i) .ii) , but (antenna 1
is selected).iii) , but (antenna 2
is selected).The probability of outage can be evaluated as the
sum of
the probabilities of the three events i), ii), and iii)(31)
The two latter components of the sum would be zero
ifinterference did not cause errors in choosing the antenna withthe
largest useful power. The three terms of (31) are nowreported
[14]i) is given by [24]:
(32)where is the outage probability at the th antenna and
isdetermined by means of (29) by properly choosing the meanpower
values related to the th antenna.ii) Prob(ii) can be derived by
means of a simplified ap-
proach: with the aim of deriving Prob(ii), we neglect
thepresence of noise; moreover, as far as interference is
consid-ered, a Rayleigh probability density function is assumed
[14].So, we have
(33)
-
VERDONE: MULTIHOP R-ALOHA FOR INTERVEHICLE COMMUNICATIONS
1005
where and denote the average value of total inter-ference at
antennas 1 and 2, respectively.iii) Prob(iii) can be evaluated in a
similar way as Prob(ii),
leading to
(34)It is worthwhile recalling that the approximated method
has
been checked in different conditions, and its validity has
beenproven in [14], [18], and [19]; when interference is due to
alow number of terminals and NLOS links (as in the
scenarioconsidered in this paper), the accuracy of the model is
shownto be very high.
REFERENCES
[1] Special issue on intelligent vehicle highway systems, IEEE
Trans.Veh. Technol., vol. 40, no. 1, 1991.
[2] Final demonstration of PROMETHEUS project, Paris, France,
Oct.1994.
[3] A. Plattner, 60 GHz mobile communications, propagation and
tech-nology requirements, in 23rd European Microw. Conf.
(EuMC93),Workshop on Mobile Communications, Madrid, Spain, Sept.
1993, pp.118120.
[4] W. Schafer, Channel modeling of short range radio links at
60 GHz formobile intervehicle communication, in Proc. IEEE Veh.
Technol. Conf.VTC91, St. Louis, MO, May 1991, pp. 314319.
[5] J. Kaltwasser and J. Kassubek, A new cooperative optimized
channelaccess for inter-vehicle communication, in Veh. Nav. Inform.
Syst. Conf.VNIS94, Yokohama, Japan, Aug. 31Sept. 2 1994, pp.
145148.
[6] T. Hatakeyama and S. Takaba, A network architecture of the
inter-vehicle packet communication system, in Veh. Nav. Inform.
Syst. Conf.VNIS94, Yokohama, Japan, Aug. 31Sept. 2 1994, pp.
159164.
[7] Y. Inoue and M. Nakagawa, MAC protocol for inter-vehicle
com-munication network using spread spectrum technique, in Veh.
Nav.Inform. Syst. Conf. VNIS94, Yokohama, Japan, Aug. 31Sept. 2
1994,pp. 149152.
[8] CEPT T/R 22-04, Harmonization of frequency bands for road
transportinformation systems (RTI).
[9] O. Andrisano, M. Chiani, V. Tralli, and R. Verdone, Impact
of cochan-nel interference on vehicle to vehicle communications at
millimiterwaves, in Proc. IEEE Int. Conf. Commun. Syst. (ICCS92),
Singapore,Nov. 1992, pp. 924928.
[10] O. Andrisano, M. Chiani, M. Frullone, C. Moss, and V.
Tralli, Mil-limeter wave short range communications for advanced
transport telem-atics, in ETT European Trans. Telecommun. Conf.,
JulyAug. 1993,pp. 403414.
[11] O. Andrisano, D. Dardari, and R. Verdone, Code division and
timedivision multiple access networks for vehicle-to-vehicle
communicationsat 60 GHz, in Proc. IEEE Veh. Technol. Conf. VTC94,
Stockholm,Sweden, June 710, 1994, pp. 18591863.
[12] W. Kremer, D. Hubner, S. Hoff, T. P. Benz, and W. Shafer,
Computer-aided design and evaluation of mobile radio local area
networks in
RTI/IVHS environments, IEEE J. Select. Areas Commun., vol. 11,
no.3, pp. 406421, 1993.
[13] R. Hager, R. Mathar, and J. Mattfeldt, Intelligent cruise
control andreliable communication of mobile stations, IEEE Trans.
Veh. Technol.,vol. 44, no. 3, pp, 443448, 1995.
[14] R. Verdone, Outage probability analysis for short range
communica-tion systems at 60 GHz in ATT urban environments, this
issue, pp.10271039.
[15] T. Liu, J. A. Sylvester, and A. Polydoros, Performance
evaluation ofR-ALOHA in distributed packet radio networks with hard
real-timecommunications, in Proc. IEEE Veh. Technol. Conf. VTC95,
Chicago,IL, July 2528, 1995, pp. 554558.
[16] W. Zhu and B. Walke, A high precision TDMA frame/slot
synchroniza-tion protocol based on a radio clock signalnew results,
in PROCOMConf., Stuttgart, Germany, 1991.
[17] W. Kremer, Vehicle density and communication load
estimation inmobile radio local area networks (MR-LANs), in Proc.
IEEE Veh.Technol. Conf. VTC92, Denver, CO, May 1992, pp.
698704.
[18] M. Chiani, V. Tralli, and R. Verdone, Outage and spectrum
efficiencyanalysis in microcellular systems, in Proc. IEEE Veh.
Technol. Conf.VTC93, Secaucus, NJ, May 1993, pp. 598601.
[19] G. Mazzini and R. Verdone, Semi-analytical evaluation of
outage prob-ability for pico and microcellular mobile radio systems
with correlatedshadowing, Electron. Lett., vol. 29, no. 13, pp.
12141215, June 1993.
[20] R. Verdone, Sullefficienza spettrale nelle reti
microcellulari e picocel-lulari, Ph.D. dissertation, Univ. Bologna,
Bologna, Italy, Mar. 1995.
[21] O. Andrisano, M. Chiani, M. Frullone, V. Tralli, and C.
Moss, Propaga-tion effects and countermeasures analysis in
vehiclevehicle communi-cation at millimeter waves, in Proc. IEEE
Veh. Technol. Conf. VTC92,Denver, CO, May 1992, pp. 312316.
[22] R. Verdone, Time and frequency selectivity effects in
vehicle-to-vehiclecommunications at 60 GHz, in Proc. IEEE Veh.
Technol. Conf. VTC94,Stockholm, Sweden, June 710, 1994, pp.
17801784.
[23] D. Dardari, A general approach to the evaluation and
characterizationof packet radio networks performance, to be
published.
[24] W. C. Jakes, Microwave Mobile Communications. New York:
Wiley,1974.
[25] V. Wong and C. Leung, Effect of Rayleigh fading in a
multihop mobilepacket radio network with capture, IEEE Trans. Veh.
Technol., vol. 44,no. 3, pp. 630637, 1995.
[26] R. Verdone, Performance evaluation of R-ALOHA for
inter-vehiclecommunications at millimeter waves, in Proc. IEEE
Personal, In-door, and Mobile Radio Communications Conf. 1996
(PIMRC96), pp.658662.
Roberto Verdone (M95) was born in Bologna, Italy, on August 6,
1965.He received the Dr. Ing. degree in electronic engineering
(with honors)and the Ph.D. degree in electronic engineering and
computer science fromthe University of Bologna, Bologna, in March
1991 and October 1995,respectively.Since 1994, he has been a
Lecturer in Telecommunications at the University
of Bologna. Since 1996, he has been a Researcher at CSITE-CNR
(ResearchCenter for Informatics and Telecommunication Systems of
the NationalResearch Council). His research activity is concerned
with digital modulation,cellular and mobile systems, multiple
access, and spectrum efficiency. Part ofhis work is dedicated to
intervehicle and vehicle-to-infrastructure communica-tions at
millimeter waves. He was involved in the European research
programPROMETHESUS and the national research program TELCO.