Power-Efficient Routing Mechanism
for ODMA Systems
Ray-Guang Cheng, Member, IEEE, Shin-Ming Cheng, and Phone Lin, Senior Member, IEEE
Abstract—Opportunity driven multiple access (ODMA), a
cel-lular multihop method proposed for Universal Mobile Telecom-munications Systems, potentially allows reduction in power consumption of user equipment (UE), extending Node B’s cov-erage and supporting higher user data rate. However, ODMA requires extra power for discovering relaying nodes and intro-duces additional transmission latency in data transfer. This paper offers enlightenment to these ODMA implementation problems. A power-efficient routing (PER) mechanism is proposed to iden-tify a minimum-power path for ODMA communication. Prior to the route (or path) discovery, the PER mechanism utilizes an analytical solution to estimate the total power and number of intermediate UEs required in the minimum-power path. With the estimation, route discovery procedures originating from nonat-tainable ODMA requests can be prevented. For those atnonat-tainable ODMA requests that require a route discovery procedure to locate intermediate UEs, the PER mechanism further provides a method to set the transmission power and maximum hop count. Hence, the power consumption of each UE during route discovery is sig-nificantly reduced. Simulation results coincided with the analysis, and the results demonstrate the performance improvement of PER over dynamic source routing.
Index Terms—Opportunity driven multiple access (ODMA),
power-efficient routing (PER).
I. INTRODUCTION
O
PPORTUNITY driven multiple access (ODMA) is an ad hoc multihop relaying protocol [1] considered by the third-generation partnership project (3GPP) working group [2]. Although it has now been dropped to achieve a finalized standard as a result of concerns over complexity, battery life of users on standby, and signaling overhead issues, ODMA remains an attractive prospect for future mobile communication systems [3]. The advantages of ODMA include 1) potentially reduced transmission power; 2) possibly enhanced coverage; 3) increased capacity under certain circumstances; and 4) a Manuscript received January 7, 2005; revised August 2, 2005, November 3, 2005, and November 11, 2005. This work was supported by the Chung-Shan Institute of Science and Technology, Taiwan, under Contract BC93203P; and in part by the National Science Council (NSC), Taiwan, R.O.C., under Contracts NSC 94-2219-E-011-005, NSC 94-2213-E-002-083, NSC 94-2213-E-002-090, and NSC 94-2627-E-002-001; the Ministry of Economic Affairs (MOEA), Taiwan, R.O.C., under Contract 93-EC-17-A-05-S1-0017; the Computer and Communications Researches Labs/Industrial Technology Research Institute (CCL/ITRL); Chunghwa Telecom Labs; Telcordia Applied Research Center; the Taiwan Network Information Center (TWNIC); and Microsoft Corporation, Taiwan, R.O.C. The review of this paper was coordinated by Dr. W. Zhuang.R.-G. Cheng is with the Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan, R.O.C. (e-mail: [email protected]).
S.-M. Cheng and P. Lin are with the Department of Computer Science and Information Engineering, National Taiwan University, Taipei 106, Taiwan, R.O.C. (e-mail: [email protected].; [email protected]).
Digital Object Identifier 10.1109/TVT.2006.877457
greater tradeoff between quality-of-service (QoS) and capacity in the extended coverage areas [3], [4].
In ODMA, user data are exchanged between a sending mobile station [also known as user equipment (UE) in Universal Mobile Telecommunications System (UMTS)] and the base station (called Node B in UMTS) by being relayed through other intermediate UEs. The sending UE should establish a path through the intermediate UEs to Node B prior to data exchange, which introduces additional signaling overhead and thus results in extra power consumption for certain UEs. Hence, a good routing mechanism with low signaling overhead would be es-sential while realizing ODMA. Unlike mobile ad hoc networks (MANETs) [5] where communication generally occurs be-tween any pair of nodes through several mobile relaying nodes, a sending UE in ODMA always exchanges data with Node B by utilizing nonmobile or low-mobility intermediate UEs [4]. Moreover, most of the nodes in MANET cannot directly com-municate with each other due to the limited transmission power. However, all UEs in ODMA can communicate with Node B directly. Hence, existing routing methods proposed for MANET may not be directly applicable to ODMA.
Several power-aware routing methods [6]–[10] have been proposed for MANET and ODMA cellular networks. Most of the proposed methods are developed out of the dynamic source routing (DSR) protocol [11] and the ad hoc on-demand distance vector (AODV) routing protocol [12]. In DSR and AODV, the source node initiates a route discovery procedure by flooding a route request (RREQ) packet its surrounding nodes. The RREQ is always forwarded by intermediate nodes until the destination node is reached. The destination node sends back a route reply (RREP) packet carrying the power metrics of the selected path(s) to the source node. A minimum-power path is then iden-tified based on the collected metrics. The power consumption of nodes in MANET was first considered by Singh et al. [6] in their routing method. Chang and Tassiulas [7] considered the residual power of UEs in their energy-efficient routing al-gorithm. Rodoplu and Meng [8] proposed a position-based routing method for mobile wireless networks. This method con-structed a position-based sparse graph for all communication links connecting mobile nodes and then derived a minimum-power routing topology from the graph. Wattenhofer et al. [9] proposed a distributed topology control algorithm for MANET. Using the directional antenna technology, each UE constructed a communication graph, removed the nonefficient edges from the graph, and derived a minimum-power routing topology. The Vodafone Group [10] proposed an ODMA routing procedure where the given local and end-to-end connectivity information was utilized to construct the routing path.
enable the discontinuous reception (DRX) function. With DRX, a UE is in sleep mode for most of the time to save power and pe-riodically wakes up to gather system information. Hence, each UE in a mobile cellular network cannot have up-to-date infor-mation of other UEs because all inforinfor-mation would be obsolete after returning to sleep mode. This assumption may be relieved by employing reactive routing approaches [13]. However, exist-ing reactive routexist-ing approaches rely on a route discovery proce-dure to obtain the other UEs’ information. Hence, some routing control messages may be wasted on processing nonattainable ODMA requests (i.e., the power or latency requirement for those requests cannot be attained by utilizing the ODMA tech-nology). The second assumption is that the extra power used by RREQ signaling is ignored. Therefore, RREQ in MANET is always flooded among UEs with UE’s maximum transmission power and without hop count limitation. UE’s transmission power can be up to several watts in a mobile cellular network, which cannot be neglected. The third assumption is that the power metric only considers the path loss between two adjacent UEs but neglects the power consumed by UEs’ receivers.
This paper presents a power-efficient routing (PER) mechanism and identifies parameters required to discover a minimum-power path for ODMA communication. Different to the existing reactive routing approaches, the PER mechanism utilizes an analytical solution to estimate the total power (i.e., including the power consumed by UEs’ receivers) and number of intermediate UEs required in the minimum-power path prior to route discovery. With the prediction, route discovery proce-dures originating from nonattainable ODMA requests can be prevented. For those attainable ODMA requests that require a route discovery procedure to locate intermediate UEs, the PER mechanism further provides a method to set the transmission power and maximum hop count when forwarding RREQ. With these settings, the power consumption of each UE during the route discovery can be significantly reduced.
The rest of this paper is organized as follows. Section II proposes the PER mechanism and discusses its key parame-ters and the effect of the parameparame-ters on system performance. Section III investigates the performance of the PER mechanism via numerical analysis and simulation. Conclusions are finally drawn in Section IV.
II. PER MECHANISM
A time division duplex (TDD)-ODMA network [3] com-prising of Node B and several nonmobile ODMA-enabled UEs, which are identified by their user-specific identities (ODMA_IDs), is considered herein. It is assumed that Node B may allocate dedicate timeslots for the ODMA commu-nication to minimize the power warfare problem [3] among ODMA and non-ODMA UEs. To simplify our description, we use the term “UE” to denote an ODMA-enable UE. In an ODMA transmission, the UEs are categorized into three
packets between the SendingUE and Node B. Note that UEs that do not have sufficient residual power may optionally dis-able some ODMA functionalities to reduce unnecessary power consumption.
This study considers three power consumption modes of the UE, including sleep (SLP), receive (RX), and transmit (TX). In SLP mode, the UE consumes the least amount of power for running a timer. In RX mode, the receiver is turned on, and the UE can receive data from other UEs and Node B. In TX mode, the transmitter is turned on, and the UE can adjust its transmission power while transmitting data. The parameters used in the PER mechanism are defined as follows.
• Pref and αPref are the minimum and maximum powers
consumed by the UE in TX mode, respectively. βPref is
the average power consumed by the UE in RX mode. γPref
is the average power consumed by the UE in SLP mode. α, β, and γ are constant numbers and with the relationship α >1 > β γ > 0 [1].
• PTX_RDPis the transmission power consumed by the UE
when forwarding RREQ in the “path discovery phase.” • Nmax is the maximum hop count that an RREQ can
traverse in the “path discovery phase.” Noptis the number
of RelayUEs required in an optimal path. The optimal path exists when the RelayUEs can be found at any location within a cell.
• Ptotal,iis the total power required by the ith path
discov-ered in the “path discovery phase.” Poptis the total power
required in the optimal path. Note that Ptotal,i≥ Popt.
• Pini is the transmission power consumed by the SendingUE to send the ODMA service request.
The PER mechanism consists of three phases, namely 1) access phase; 2) path discovery phase; and 3) path setup phase. In access phase, the SendingUE adjusts its transmission power to Piniand sends an ODMA service request carrying Pini
to Node B. Node B can predict Poptand Noptbased on Pini. By
using the predicted Poptand Nopt, Node B checks whether the
ODMA request is attainable or not. For nonattainable ODMA requests, Node B simply terminates the PER procedure by replying a rejection message to the SendingUE. For attainable ODMA requests, Node B further derives PTX_RDPand Nmax,
and sends a confirmation message carrying PTX_RDP and Nmax to the SendingUE. In the path discovery phase, similar
to DSR [11], the SendingUE broadcasts an RREQ through the ith path to Node B to collect Ptotal,i. In this phase, each BackerUE floods the RREQ with transmission power PTX_RDP
and discards the RREQ that exceeds the hop count limitation Nmax. Based on the collected Ptotal,i, Node B can identify
the minimum-power path. As an option, Node B may refuse the ODMA request ifminiPtotal,i Popt. In the “path setup
phase,” Node B sends an RREP packet along the identified path to configure the RelayUEs.
Fig. 1. Colinear network topology consisting ofN + 2 colinear nodes, UE1, . . . , UEN+1, and Node B. Note that the proposed PER differs from DSR in the
fol-lowing respects. First, PER can predict Popt before the route
discovery. Second, the transmission power utilized to forward the RREQ is αPrefin DSR but is PTX_RDPin PER(αPref PTX_RDP). Third, the hop count limitation for RREQ is infinite
in DSR but is Nmax in PER. The derivations of PTX_RDP, Nmax, and Piniare elucidated next.
We investigate Poptby considering a colinear network
topol-ogy as shown in Fig. 1, where Node B, N RelayUEs, and SendingUE (i.e., U E1) are located along a line. For the sake
of simplicity, RelayUEs are numbered in order and denoted as U Ej, where j= 2, . . . , N + 1. Let d be the distance between SendingUE and Node B. Assume that the UE density in a cell is sufficiently high such that a UE can be found at any location along the line. Let the continuous random number ˜dj denote the distance between U Ej and U Ej+1. During ODMA communication, the UEs are operating in TX and RX modes. In TX mode, the transmission power required by a UE depends on the radio channel condition. Typically, the radio channel con-dition is characterized by a large-scale propagation model1and
a small-scale propagation model2[14]. Rodoplu and Meng [8]
have proven that a minimum-power network design is funda-mentally the same as that which considers only the path loss. Hence, only a path loss propagation model was considered in the analysis and simulation. The path loss model with the following parameters is used herein: a power law attenuation factor n(4 ≥ n ≥ 2), antenna gain of a UE’s transmitter (re-ceiver) Gt(Gr), the wavelength of the modulated signal λ, the system loss factor L(L ≥ 1) [14], and the power required by the UE to correctly decode a message Pd. Note that Pdcould be properly set by considering the effects of shadowing and fast fading in the implementation. With the characteristics, we have the following lemma.
1A large-scale propagation model is utilized to predict the mean signal power
for a relatively long transmitter–receiver separation. The path loss and the shadowing effect are considered.
2Small-scale propagation model characterizing the rapid fluctuations of
the received signal strength over a very short distance. Delay spread due to multipath and Doppler effects is considered.
Lemma 1: The total power (Popt) and the number of RelayUEs (Nopt) required in the optimal ODMA path are
shown in (1) and (2), respectively, at the bottom of the page. In (1) and (2), k= ((4π)2L/G
tGrλ2)Pd.
Proof: Denote PTX,jand PRX,jas the powers of U Ejin TX mode (where αPref≥ PTX,j ≥ Pref) and RX mode (where PRX,j= βPref), respectively. The power transmitted by U Ej and received by U Ej+1, denoted as Pr,j+1( ˜dj), is obtained by applying the Friis free space equation [14]
Pr,j+1( ˜dj) = PTX,jGtGrλ2 (4π)2d˜n jL ∆ = 1 k0 PTX,jd˜−nj (3)
where k0= (4π)2L/GtGrλ2. For successful reception,
Pr,j+1( ˜dj) should not be less than Pd. Hence
PTX,j= k0d˜njPr,j+1( ˜dj) ≥ k0d˜njPd ∆= k ˜dnj (4) where k= k0Pd. The variable Ptotal,iis obtained by summing
the power required by all transmitters and receivers of the SendingUE and RelayUEs. That is,
Ptotal,i= N +1 j=1 PTX,j+ N +1 j=2 PRX,j = N +1 j=1 k ˜dn
j + NβPref, for k ˜dnj > Pref
(N + 1)Pref+ NβPref, for Pref≥ k ˜dnj.
(5)
First, we consider the case that PTX,j> Pref, that is,
˜ dj >
[n]Pref
k . (6)
By taking the expectation on both sides of (6) and replacing E[ ˜dj] with d/(N + 1), we obtain dn k Pref − 1 > N ≥ 0. (7) Popt= min i Ptotal,i|N =Nopt = k dn
(Nopt+1)n−1+ NoptβPref, for0 < Nopt< n
k Prefd− 1
(Nopt+ 1 + Noptβ)Pref, for Nopt ≥ n
k Prefd− 1 (1) Nopt= n k Prefd− 1, , if n k Prefd− 1 + βPref > kd n n k Prefd n−1 n k Prefd− 1 , otherwise (2)
for dn
k Pref
− 1 > N ≥ 0. (8) The lower bound of (8) is obtained by varying ˜dj, that is,
min
i Ptotal,i= Ptotal,i|d˜j= ¯dj, forj=1,...,N +1. (9) The optimal distance between adjacent nodes ¯djthat results in the minimum Ptotal,i is obtained by(∂/∂ ˜dj)Ptotal,i= 0, for j= 1, . . . , N + 1. Thus ¯ d1= ¯d2= · · · = ¯dN +1= d − N i=1 ¯ di= d N+ 1. (10) Equation (10) demonstrates that, for a given N , the lower bound is achieved if the distances between any two adjacent RelayUEs are equal. Under this condition, the transmission power re-quired by each RelayUE to reach its neighboring RelayUE is a constant, which is denoted as P0, where
P0= k ˜dni|d˜i= ¯di= k d N+ 1 n . (11)
From (8) and (11),miniPtotal,iis obtained by
min i Ptotal,i= (N + 1)P0+ NβPref = k d n (N + 1)n−1 + NβPref for 0 < N < n k Pref d− 1. (12) Equation (12) is a monotonically decreasing function of N because(d/dN)Pt<0, for n ≥ 2 and N < n (k/P
ref)d − 1.
Since N should be an integer, Poptis obtained when
N = Nopt= n k Pref d− 1 . (13)
Now consider the case that PTX,j ≤ Pref, or equivalently, N ≥
n
k/Prefd− 1. From (5), miniPtotal,iis obtained by
min i Ptotal,i= (N + 1 + Nβ)Pref for N ≥ n k Pref d− 1. (14) Equation (14) is a monotonically increasing function of N . Since N should be an integer, Poptis located when
N = Nopt= n k Pref d− 1 . (15)
Combining (13) and (15), we obtain (2). And, Poptgiven in (1)
is obtained by combining (12) and (14).
parameters, the only unknown factor is d. In mobile cellular networks, UE normally utilizes an open-loop power control mechanism [14] to estimate d. Let PBCH and Pavg be the
broadcast channel (BCH) power transmitted by Node B and the average power received by SendingUE, respectively. In UMTS, PBCHis a constant and is periodically broadcasted by Node B.
Hence, SendingUE can estimate d from (4), that is,
d= n
PBCH kPavg
. (16)
From (4) and (16), the initial transmission power used by SendingUE to send the ODMA service request to Node B Pini
is given by
Pini = kdn = PBCH
Pavg
. (17)
Lemma 1: suggests that, with Nmax= Nopt+ 1 and PTX_RDP= P0= k(d/Nmax)n, an optimal path in a colinear
network topology is obtained given sufficiently high UE den-sity. For a cellular network with low UE density, the optimal path may not be found. To solve this problem, RelayUE must increase PTX_RDP to find another RelayUE in its
neighbor-hood. Therefore, the minimum-power path can still be obtained if PTX_RDP= δP0 (i.e., αPref/P0≥ δ ≥ 1) is applied. Note
that, under this condition, the total power required by the minimum-power path is not less than Popt.
Normally, RelayUEs are located between SendingUE and Node B. As demonstrated in Fig. 2, BackerUEs in the region where the two circles overlap (both solid circles centered at Node B and SendingUE have the same radius Pini) could
be possible RelayUE candidates. Hence, in PER, only these BackerUEs, rather than all BackerUEs in the entire cell, should forward RREQ during route discovery. These BackerUEs can be identified easily because they can receive both the ODMA service request and the confirmation from SendingUE and Node B, respectively.
Figs. 2 and 3 show a general network topology and the message flows employed to demonstrate a scenario of the PER mechanism, respectively. In this scenario, U E1 is the Send-ingUE, U Ejs, for j= 2, . . . , 12, are BackerUEs, and Nopt = 1
is assumed. As shown in Fig. 2, U E11 cannot receive the
ODMA service request from U E1, and U E12 cannot receive
the confirmation from Node B; hence, U E11and U E12
auto-matically enter the SLP mode after timeout. The RREQ mes-sage traversing along U E1−UE6−UE7 is discarded by U E7
because Nmax is reached. Not otherwise specified, messages
are carried through the logical channels specified in parenthesis in Fig. 3 (i.e., ORACH denotes the ODMA random access channel [2]). The three phases of the PER mechanism are described as follows.
Fig. 2. Network topology illustrating the PER mechanism.
Fig. 3. Message flow of the PER mechanism.
A. Access Phase
Step 1) Prior to communicating with Node B, the SendingUE U E1 measures Pavg, adjusts its
trans-mission power to Pini, and then sends RRC_
Connection_Req [2] carrying Pinito Node B.
Step 2) Upon receiving the RRC_Connection_Req mes-sage, Node B rejects the request if the request is nonattainable. Otherwise, Node B derives PTX_RDP
and Nmax, adjusts its transmission power to Pini,
and acknowledges ODMA_Relay_Prepare carrying PTX_RDPand Nmaxto U E1.
B. Path Discovery Phase
In the path discovery phase, the SendingUE adjusts its transmission power to PTX_RDP and floods an RREQ (i.e.,
Fig. 4. Message flow of the PER mechanism link management functions.
ODMA_Relay_Req) the surrounding BackerUEs. The RREQ carries three parameters: namely 1) SID; 2) RoutingList; and 3) Pacc,j. The SID is the ODMA_ID of the SendingUE and
is utilized to identify a specific ODMA connection request; the RoutingList contains ODMA_IDs of UEs that comprise the specific path; and the Pacc,jis the accumulated power required
for the path from SendingUE to U Ej.
Step 3a) U E1 sends ODMA_Relay_Req carrying (SID=
1, RoutingList = NULL, Pacc,1= 0) to its
neighboring UEs, and U E7 updates the
accumu-lated power by Pacc,7= Pacc,1+ PTX,1+ PRX,7
= Pacc,1+ max(PTX_RDP− Pr,7, Pref) + βPref (18)
where Pr,7 is the power of U E1’s ODMA_
Relay_Req measured at U E7.
Step 4a) U E7 forwards the RREQ carrying (SID= 1, RoutingList= 7, Pacc,7) to Node B. Node B updates the total power Ptotal,iof this first path by Ptotal,1= Pacc,7+ PTX,7
= Pacc,7+ max(PTX_RDP− Pr,NodeB, Pref) (19)
where Pr,NodeB is the power of U E7’s
ODMA_Relay_Req measured at Node B. Note that the power used by Node B’s receiver is a common factor for all paths and thus is not considered in calculating Ptotal,i.
Step 3b) U E6 receives the RREQ from U E1, updates the
triplet for this second path, and forwards the RREQ to U E7.
Step 4b) U E7discards the RREQ because Nmaxis reached.
C. Path Setup Phase
Step 5) Node B determines the minimum-power path, which has the least Ptotal,iamong all discovered paths, and
identifies U E7as the RelayUE from the RoutingList.
Step 6) Node B sends an RRC_Connection_Setup [2] to U E7carrying the ODMA traffic channel (ODTCH)
and ODMA control channel (ODCCH) allocations [2]. The remaining BackerUEs whose ODMA_ID are not on the RoutingList move to the SLP mode. Step 7) The ODMA communication path is established. The
established communication path may be broken if mobility UEs are further considered. In such a mo-bile environment, Node B may repeat Steps 5) to 7) to create one or more backup communication paths to the SendingUE and enable link management func-tions for managing these paths. In the implemen-tation, PER employs a well-known sliding-window scheme with a stop-and-wait automatic retransmis-sion request (ARQ) mechanism to control data flow and retransmit error packets between adjacent RelayUEs. The same network topology shown in Fig. 2 is utilized to demonstrate a scena-rio of link management functions. In this sce-nario, U E1−UE7–NodeB is the primary path and U E1−UE6–NodeB is the backup path. Fig. 4 shows
the message flow of the link management functions. In this scenario, the sliding-window size Wmaxis 2,
and timers Ttand Tsware required by UEs to control packet retransmission and by Node B to control path switching, respectively.
D. Flow Control and Error Control
Step 1) U E1 starts its Ttand transmits packet A1 through the primary path to U E7.
Step 2) U E7sends an acknowledgment to U E1denoting the
successful reception of A1; then, U E1 resets and
stops its Tt.
Step 3) U E7 starts its Tt and forwards A1 to Node B. Node B resets and starts Tsw whenever it correctly receives a new packet.
Step 4) U E7 resets and stops its Ttafter receiving the ac-knowledgment that A1 was received.
Step 5) U E1starts its Ttand transmits a packet A2 to U E7.
A2 is lost.
Step 6) U E1retransmits A2 after the expiry of its Tt. A2 is lost and Ttexpires again. Since A2 has transmitted
Wmaxtimes, the retransmission is stopped.
E. Switch to the Backup Path
Step 7) After Tsw expires, Node B switches to the backup path by sending ODMA_Path_Switch to U E6and U E1to activate the backup path.
Step 8) U E1starts its Ttand transmits the last unac-knowledged packet A2 over the activated path to U E6.
Step 9) U E1 resets and stops Tt after receiving an acknowledgment from U E6.
Step 10) U E6starts its Ttand forwards A2 to Node B; Node B resets and starts its Tswafter receiving a packet from U E6.
Step 11) U E6resets and stops its Ttafter receiving an acknowledgment from Node B.
Steps 12)–15) U E1successfully transmits A3 to Node B.
III. NUMERICALRESULTS
Simulations were conducted to verify the effectiveness of the proposed PER mechanism. The load balancing capability of ODMA was not investigated herein. Hence, a single cell with 50–500 nonmobile UEs was considered. All UEs were assumed to be uniformly distributed within a hexagonal cell with ra-dius 2500 m. The constants used herein are listed as follows: λ= 15.78 cm, Gt= Gr= 1, L = 1, k0= 6334, α = 20, β= 0.1, . . . , 0.9, n = 2, Pref= 20 mW, Pd= 10−8mW, d= 2100 m, and δ = 2. Each sample during the simulation was obtained by averaging the outcomes from106identical exper-iments. Both DSR and PER were simulated. The DSR was chosen as a benchmark because it can explore all paths and identify the minimum-power path in a cell. In the simulation, both DSR and PER found the same minimum-power path,
Fig. 5. Total power required by the path for various UE densities andN.
Fig. 6. Total power required by the path for variousN and β.
but with different signaling overheads. Hence, the minimum-power path discovered by DSR was not specifically identified in Figs. 5 and 6. In Figs. 5 and 6, the analytical results are denoted with lines, while the simulation results are presented with symbols.
The accuracy of the analysis was first verified by simulation. In Fig. 5, the total power required by the minimum-power path (i.e., Ptotal, where Ptotal miniPtotal,i) for various UE
densities and the number of RelayUEs (i.e., N ) were shown, in which β= 0.5 was assumed. Lemma 1 obtained Nopt= 3
and Popt= 110 mW. Note that for d = 2100 m, SendingUE
required 279 mW to transmit data directly to Node B without using ODMA. Simulation results showed estimation errors for low UE densities (Fig. 5). However, the estimation error was considerably reduced when the UE density was larger than 5 × 10−6 UEs/m2. This finding was a result of the high UE
Fig. 7. Signaling cost of DSR and PER. (a) Total number of RREQ messages. (b) Total power consumed by RREQ messages.
could not be found at expected locations, and therefore, the lower bound was not achieved.
In the following two examples, the UE density is fixed to 2 × 10−5UEs/m2. Fig. 6 showed P
totalfor various N s and βs.
From Lemma 1, it can be derived that Nopt= 2 for β = 0.7
and β= 0.9, and Nopt= 3 for β = 0.1, β = 0.3, and β = 0.5;
each derived Noptcoincided with the simulation results shown
in Fig. 6. Fig. 6 demonstrated that for a fixed N , a decreased β resulted in a lowered Ptotal since a low power is required
by the receiver of each RelayUE. For a given β, Ptotal was
first decreased and then increased when N was increased from 1 to 6. The rationale for the variation of Ptotal is described
as follows. Increasing N meant to add new RelayUEs in the path. Since these new RelayUEs consume extra power, it is not valuable to reduce Ptotalby increasing the number of RelayUEs
unlimitedly, particularly for those RelayUEs that have high β. In other words, using RelayUEs closer than1/(Nopt+ 1)
together results in greater overall power consumption since the savings in TX power from using smaller hops is lost given that nothing less than Pref can be used. Lemma 1 proved that the
minimum Ptotal was obtained if Nopt RelayUE was utilized
in a path. For N < Nopt, increasing N implied a decrease
in the distance between two adjacent RelayUEs; hence, the transmission power of existing RelayUEs was reduced. How-ever, the cost was the extra power consumption introduced by new RelayUEs. In the region of N < Nopt, Ptwas decreased because the power required by new RelayUEs is less than the power reduced by existing RelayUEs. However, in the region of N > Nopt, reducing the distance between two adjacent RelayUEs did not further reduce the transmission power of each RelayUE because the transmission power was bounded by Pref;
therefore, Ptotalwas monotonically increased.
As mentioned earlier, both DSR and PER were able to locate the same minimum-power path; however, their signal costs were substantially different. In DSR, the UEs flood the RREQ over the entire cell with transmission power αPref.
However, in PER, only selected BackerUEs flood the RREQ with transmission power δP0. Fig. 7 shows the signaling cost of
DSR and PER. The number of RREQs, (i.e., denoted as Nsignal)
and the total power consumed by the RREQs (i.e., denoted as Psignal) were investigated and illustrated in Fig. 7(a) and (b),
respectively. In this example, β= 0.5, and δ = αPref/P0and δ= 2 were used in PER1and PER2, respectively. That is, both
PER1 and DSR used UE’s maximum transmission power to
flood the RREQ. As shown in the figures, the proposed PER mechanism dramatically reduced Nsignal and Psignal because,
in PER, fewer BackerUEs were allowed to forward the RREQ. The figures also demonstrated that a small δ results in a small Nsignal and Psignal. However, reducing Nsignal by lowering δ increased the risk of locating no path during the route discovery, particularly for those networks with low UE density. Since the optimization of δ is not essential for the effectiveness of the PER mechanism, its optimization will be the subject of future work.
IV. CONCLUSION
This paper presents a PER mechanism for ODMA cellular networks. In contrast to previous routing approaches, the pro-posed PER mechanism can estimate the power consumption of, and the number of relay nodes for, an optimal path without information from the other nodes. With the estimation, route discovery procedures originating from nonattainable ODMA requests can be prevented. The PER mechanism further pro-vides attainable ODMA requests, a method to set the trans-mission power and maximum hop count to reduce the power consumption of each UE during the route discovery. The ef-fectiveness of the proposed method is shown both theoretically and via simulation. Simulation results demonstrate that, with carefully chosen parameters, the PER mechanism can identify the minimum-power path with relatively low signaling cost compared to that of DSR.
TABLE I PARAMETERDEFINITION
APPENDIX
The definition of parameters involved in the analysis is summarized in Table I.
ACKNOWLEDGMENT
The authors would like to thank anonymous reviewers for their valuable comments that helped improve the quality of the paper.
REFERENCES
[1] H. Karl, “An overview of energy-efficiency techniques for mobile com-munication systems,” Telecomcom-munication Networks Group, Technical University Berlin, Berlin, Germany, Tech. Rep. TKN-03-017, Oct. 2003. AG Mobikom WG7.
[2] “Opportunity Driven Multiple Access,” Tech. Rep. 25.924, Dec. 1999. 3GPP, version 1.0.0.
[3] T. Rouse, I. Band, and S. McLaughlin, “Congestion-based routing strate-gies in multihop TDD-CDMA networks,” IEEE J. Sel. AreasCommun., vol. 23, no. 3, pp. 668–681, Mar. 2005.
[4] P. Lin, W.-R. Lai, and C.-H. Gan, “Modeling opportunity driven multiple access in UMTS,” IEEE Trans. Wireless Commun., vol. 3, no. 5, pp. 1669–1677, Sep. 2004.
[5] A. J. Goldsmith and S. B. Wicker, “Design challenges for energy-constrained ad hoc wireless networks,” IEEE Wirel. Commun., vol. 9, no. 4, pp. 8–27, Aug. 2002.
[6] S. Singh, M. Woo, and C. S. Raghavendra, “Power-aware routing in mobile ad hoc networks,” in Proc. 4th Annu. ACM/IEEE Int. Conf. Mobile
Comput. and Netw. (MobiCom), 1998, pp. 181–190.
[7] J.-H. Chang and L. Tassiulas, “Energy conserving routing in wireless
ad-hoc networks,” in Proc. IEEE 19th Annu. Joint Conf. IEEE Comput. and Commun. Soc. (INFOCOM), 2000, vol. 1, pp. 22–31.
[8] V. Rodoplu and T. H. Meng, “Minimum energy mobile wireless net-works,” IEEE J. Sel. AreasCommun., vol. 17, no. 8, pp. 1333–1344, Aug. 1999.
[9] R. Wattenhofer, L. Li, P. Bahl, and Y. M. Wang, “Distributed topol-ogy control for power efficient operation in multihop wireless ad hoc networks,” in Proc. IEEE 20th Annu. Joint Conf. IEEE Comput. and
Commun. Soc. (INFOCOM), 2001, vol. 3, pp. 1388–1397.
[10] “ODMA Routing With Procedures for Mobile Originated Calls, Mobile Terminated Calls, and Location Update,” Tech. Rep. Tdoc TSGR2#2(99) 179, 1999. 3GPP TSG-RAN WG2.
[11] D. Johnson and D. Maltz, “Dynamic source routing in ad hoc wire-less networks,” in Mobile Computing. Norwell, MA: Kluwer, 1996, pp. 153–181.
[12] C. E. Perkins, E. M. Belding-Royer, and S. Das, ad hoc On-Demand
Distance Vector (AODV) Routing, Jul. 2003. RFC 3561.
[13] A. M. Safwat, H. S. Hassanein, and H. T. Mouftah, “Structured pro-active and repro-active routing for wireless mobile ad hoc networks,” in
The Handbook of ad hoc Wireless Networks. Boca Raton, FL: CRC,
pp. 233–244.
[14] T. S. Rappaport, Wireless Communications: Principles and Practices, 2nd ed. Englewood Cliffs, NJ: Prentice-Hall, 2002.
[15] M. Bhardwaj, T. Garnett, and A. P. Chandrakasan, “Upper bounds on the lifetime of sensor networks,” in Proc. IEEE ICC, 2001, vol. 3, pp. 785–790.
Ray-Guang Cheng (S’94–M’97) received the B.E.,
M.E., and Ph.D. degrees in communication engi-neering from the National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 1991, 1993, and 1996, respectively.
From 1997 to 2000, he was as a Researcher and a Project Leader with the Advance Technology Center, Computer and Communication Laboratories, Industrial Technology Research Institute (ITRI). He was involved in the designing of medium access control and radio resource management algorithms for GPRS and UMTS systems. His team was named “Top Research Team of the Year” by ITRI and he received the “Outstanding Technology Prize” from Min-istry of Economic Affairs in 2000. From 2000 to 2003, he was the Senior Man-ager of the R&D Division, BenQ Mobile System Inc. Since 2003, he has been with the Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, as an Assistant Professor. His research interests include multihop wireless networks and multimedia communications. Dr. Cheng is a member of the IEEE Communication Society and the Phi Tau Phi Scholastic Honor Society. He received the “Best Industrial-based Paper Award” from the Ministry of Education, Taiwan, in 1998.
Shin-Ming Cheng received the BSCSIE degree
from the National Taiwan University of Science and Technology, Taipei, Taiwan, R.O.C., in 2000. He is currently working toward the Ph.D. degree in the Department of Computer Science and Information Engineering, National Taiwan University of Science and Technology.
His current research interests include mobile com-puting, personal communications services, and wire-less Internet.
Phone Lin (M’02–SM’06) received the B.S. degree
in computer science and information engineering and the Ph.D. degree from the National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 1996 and 2001, respectively.
From August 2001 to July 2004, he was an Assis-tant Professor with the Department of Computer Sci-ence and Information Engineering (CSIE), National Taiwan University (NTU), Taipei, Taiwan, R.O.C. Since August 2004, he has been an Associate Profes-sor with the Department of CSIE and the Graduate Institute of Networking and Multimedia, NTU. His current research interests include personal communications services, wireless Internet, and performance modeling.
Dr. Lin is an Associate Editor of the IEEE TRANSACTIONS ONVEHICULAR
TECHNOLOGY, a Guest Editor of the IEEE TRANSACTIONS ONWIRELESS
COMMUNICATIONSspecial issue on Mobility and Resource Management, and a Guest Editor for the Association for Computing Machinery/Springer Mobile
Networksand Applicationsspecial issue on Wireless Broad Access. He is also
an Associate Editorial Member for the Wireless Communications and Mobile
Computing Journal. He has received many research awards, such as the Young
Researchers Award from the Pan Wen-Yuan Foundation in Taiwan in 2004, the K. T. Li Young Researcher Award from the ACM Taipei Chapter in 2004, the Wu Ta You Memorial Award of the National Science Council of Taiwan in 2005, and the Fu Suu-Nien Award from NTU in 2005. He is listed in Who’s
