A joint power and rate assignment algorithm for multirate soft handoffs in mixed-size WCDMA cellular systems
11
0
0
全文
(2) LIAO et al.: JPRA ALGORITHM FOR MULTIRATE SOFT HANDOFFS IN MIXED-SIZE WCDMA CELLULAR SYSTEMS. power because of serving SHO users in downlink [10], [11], and then, there would have no enough power resource to serve other NHO users in the system. When taking into account multirate services, this power-exhausting problem will become more critical. References [15]–[17] considered capacity issues in mixed-size cellular systems. References [15] and [16] focused on the reverse link, and only voice service was considered. Kishore et al. [17] concluded that uplink and downlink directions were equivalent in mixed-size cellular systems. Also, it did not consider multirate services. Note that multirate services are often regarded as highly resource-exhausting traffics and often have more volumes of traffic in downlink than that in uplink. In this paper, a novel joint power allocation and rate assignment (JPRA) algorithm is proposed for downlink multirate SHOs in mixed-size WCDMA cellular systems. We first formulate this joint PA and RA issue as a combinatorial optimization problem. Then, the JPRA algorithm is designed in a cooperative two-phased process, which is composed of the link proportional PA (LPPA) scheme in the first phase and the evolutionary computing RA (ECRA) method in the second phase. The LPPA scheme is for PA of SHOs. Unlike the SSDT scheme, LPPA is a multisite transmission mechanism, which distributes the required power in proportion to link qualities between a SHO user and all BTSs in its active set. The BTSs in the active set with better link quality will allocate more power than others with worse link qualities. This will result in power balancing among cells. Also, the ECRA method, which is based on PA of supportable transmission rates for multirate SHO users obtained by LPPA, formulates the joint PA and RA issue to be an integer and discrete optimization problem, where the sum of the allocated rate for multirate SHO users can be maximized, under a predefined total power constraint for SHOs in each cell. It is well known that conventional optimization methods can hardly cope with problems with integer and discrete variables, whereas evolutionary computing methods are very efficient for these problems to reduce the searching complexity [18]. In the meantime, a new multiquality balancing PA (MQBPA) algorithm for NHO users with multiple service rates is also developed. With the MQBPA algorithm, each BTS can allocate the power required for each NHO user based on the ratio of the user’s received interference to the link quality. Previous works for quality balancing PA technique were studied only for a single service rate with unique required signal quality [19], [20]. On the other hand, a multirate removal (MRV) algorithm is proposed to pick out a user, who consumes system resource most, to reduce its service rate or even block it when the system resource is insufficient. Several removal algorithms had been proposed [21]–[23]. Among these, link-based and received signal-strength-based removal algorithms were only suitable for single service [21], [22]; the prioritized removal algorithm [23], which is based on predefined service priority, did not adjust the service rates for users in the reverse link of a multiservice cellular system. Simulation results show that the JPRA algorithm can achieve less forced handoff termination probability by over 300% and higher system throughput by around 5.0% than the conventional SSDT scheme with best effort RA. Also, on the perspective of. 1389. Fig. 1. System operation of downlink power and rate allocations.. users, it can support excellent user satisfaction indexes (USIs) for voice and data users. Moreover, if the measurement error (ME) during the active set selection is happened, JPRA behaves less sensitive than SSDT; JPRA can further enhance the forced termination probability of SHOs by about 500% and the total throughput by 8.0% over SSDT. The remaining parts of this paper are organized as follows. Section II details the flow of the system operation and provides the design of MQBPA and MRV algorithms. In Section III, the JPRA algorithm for multirate SHOs is presented, including LPPA and ECRA algorithms. Simulation results are provided and discussed in Section IV. Finally, Section V presents some conclusions. II. S YSTEM O PERATION The system operation of downlink power and rate allocations for mixed-size WCDMA cellular systems is shown in Fig. 1. A BTS allocates power to HO users first, based on the new JPRA algorithm, and then to NHO users, which is based on the MQBPA algorithm. If the system resource is insufficient to support all users with allocated rates in their required signal qualities, an MRV algorithm is activated to release system resources by reducing the users’ service rates or even suspending the users’ transmission, and the system operation is executed again. A. System Model In the multirate WCDMA mixed-size cellular system, the received interference of user j served by BTS i, which is.
(3) 1390. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 56, NO. 3, MAY 2007. denoted by Ii,j , is Ii,j = (1 − α)Pi Li,j +. . Pk Lk,j + ηo. (1). k=i. wi,j =. where α is the orthogonality factor, Pi is the downlink total transmission power of cell i, Li,j is the link quality from cell i to user j, which includes effects of both pathloss and shadowing, and ηo is the background noise. Note that the first and second terms in (1) denote intracell and intercell interferences, respectively, in which the first term is caused by imperfect orthogonality of spreading codes. Each user is with service rate r, where voice users are with single rate r = rv and data users are with one of M kinds of data service rates, r = rd ∈ {rd1 , . . . , rdM } and rd1 < rd2 < · · · < rdM . The received bitenergy-to-noise ratio (Eb /No ) of user j with service rate r in BTS i, which is denoted by γi,j (r), must be larger than or equal to the required signal quality, which is denoted by γ ∗ (r). With allocation power qi,j (r) from BTS i to user j, γi,j (r) can be expressed as γi,j (r) =. qi,j (r) · Li,j · G(r) ≥ γ ∗ (r) Ii,j. (2). where G(r) = W/r is the processing gain and W is the frequency bandwidth. Assume there are Nv voice HO users and Nd data HO users in the system. For each SHO user h with service rate r, its received Eb /No , γh (r), can be obtained, if using the maximum ratio combining method to combine signals from all serving BTSs in active set Dh , by γh (r) =. . γi,h (r). (3). i∈Dh ∗. where rh (r) ≥ γ (r). The active set of each user h is determined according to an SHO algorithm which has SHO threshold η. If the strength difference between the received pilot signals of the original cell and a target cell is less than or equal to η, the target cell would be added as an active member in action set Dh . B. MQBPA Algorithm The MQBPA algorithm is to provide each NHO user the required signal quality of itself. Assume each service rate r has the required signal quality γ ∗ (r); denote Ci (Qi ) as the total transmission power for NHO (SHO) in cell i such that Ci + Qi = Pi ; denote Pi as the maximum total transmission power of cell i. The MQBPA algorithm assigns the NHO user j in cell i with service rate r an amount of power qi,j (r) by wi,j qi,j (r) = · Ci wi,j j∈Ui. where Ui is the set of NHO user in cell i, j∈Ui qi,j (r) = Ci ; wi,j is the weight factor of the transmission power Ci for user j in cell i, which is defined as. (4). Ii,j · γ ∗ (r). Li,j · G(r). (5). This wi,j is designed to combat the near–far effect for the WCDMA cellular system. In order to obtain the required balanced signal quality γ ∗ (r)/G(r), which is based on (2), the required transmission power of each user j in BTS i with rate r, qi,j (r), must be proportional to the ratio of Ii,j /Li,j . Thus, more (less) power would be allocated to the user with higher (lower) transmission rate r or larger (smaller) ratio of Ii,j /Li,j . In the MQBPA algorithm, each user has to transmit the value of wi,j , which can be obtained by measuring its received total interference and signal strength, to its serving BTS via the control channel. However, ME might be occurred in this stage, and it might degrade the system performance. We consider the case of that MS results in misarranging the users’ active sets in Section IV. Substituting (4) and (5) into (2), the received signal quality of the NHO user j in cell i can be yielded as Ci γ ∗ (r). γi,j (r) = wi,j. (6). j∈Ui. With the system requirement of that γi,j (r) ≥ γ ∗ (r), if Ci / j∈Ui wi,j is less than one, the total allocation power Ci of BTS i should be adjusted by tuning factor ψi , which is given by ψi =. γ ∗ (r) . γi,j (r). (7). In the following, the MQBPA algorithm is described [The MQBPA Algorithm] Step1: [Initialize] • Initialize Pi to be Pi for each cell i, and calculate Qi for SHO users in each cell i after executing JPRA algorithm. • Initialize incremental tuning value ∆ψi and the last tuning value ψi of cell i to be zero. Step 2: [Calculate wi,j ] • Calculate wi,j , based on (5), for user j in cell i. Step 3: [Calculate qi,j (r)] • Calculate Ci = Pi − Qi for NHO users in each cell i. • Calculate qi,j (r) for each NHO user j with service rate r in cell i based on (4), and qi,j (r) = min( qi,j (r), qi ). Step 4: [Calculate ψi ] • Calculate ψi for each cell i, based on (6) and (7). Step 5: [Calculate ∆ψi and save current tuning factor as ψi ] • Calculate ∆ψi = ψi − ψi for each cell i. • Save ψi as ψi for each cell i. Step 6: [Check Stop Criterion for each cell i] • IF any ψi = 1.0 or any ∆ψi > 0.01, denoting the convergence is not met, THEN.
(4) LIAO et al.: JPRA ALGORITHM FOR MULTIRATE SOFT HANDOFFS IN MIXED-SIZE WCDMA CELLULAR SYSTEMS. 1391. —Adjust the total transmission power as Pi = min (ψi × Ci + Qi , Pi ).. (8). —Goto Step 2. ELSE DONE. The proposed MQBPA algorithm will converge to a desired solution of that an effective individual PA for each users exits and its required signal quality is satisfied. The convergence characteristic of (8) can be proved, similar to the proof in [20], where the power control algorithm for the single service is standard for it satisfies three properties: positivity, monotonicity and scalability. If the solution does not exist, the MRV algorithm will be activated. C. MRV Algorithm In order to squeeze resource from a user who consumes the largest power whenever lacking of resource enough for accommodating all users, a removal algorithm is launched to pick out a user based on a removal index. In this paper, the proposed MRV algorithm, for the mixed-size WCDMA cellular system with multirate services, defines a new removal index for user j with service rate r, which is denoted by Jj (r), as Jj (r) = γ ∗ (r)/{Pi · Li,j · G(r)}, where Pi is the pilot power of BTS i, which is related to the cell size, and Pi · Li,j is equal to user j’s received signal strength from BTS i. The removal index represents the degree of system resource required to serve user j with service rate r. A user with the largest value of the removal index will consume the most system resource currently, and this user should be first considered to be removed or reduced rate. The worse the received signal strength, the higher the service rate and required signal quality are, the larger the removal index value will be. In order to provide higher priority for voice users, the proposed MRV algorithm removes system resource from data users first unless all the data users are reduced to basic service rate, which is the lowest transmission rate for data services. The flowchart of MRV algorithm is shown in Fig. 2. At first, the MRV scheme will check if all data users are with basic rate. If there exists at least one data user not with the basic rate, the MRV scheme will choose the data user with the maximum removal index. If the service rate of the selected user is with the basic rate, then the system will remove it directly, otherwise reduce its rate to the next lower service rate. If all data users are with basic rate, the system will remove the user which is with the maximum removal index.. Fig. 2. Flowchart of the MRV algorithm.. received bit-energy-to-noise ratio of each SHO user must not be less than its requirement of signal quality. We define the total throughput of multirate SHO to be the sum of the allocated rate for the multirate SHO data users, excluding the service rate of SHO voice user which is constant. Also, for clarity, denote the allocation rate r for SHO data user h as (rd )h , (rd )h ∈ {rd1 , . . . , rdM }, and the rate vector for all multirate SHO data users by R = [(rd )1 , . . . , (rd )h , . . . , (rd )Nd ]. Then, the optimization problem is mathematically formulated by ∗. R = arg max R. N d . (rd )h. (9). h=1. subject to constraints for r ∈ {rv , rd1 , . . . , rdM } Nv +Nd . i , qi,h (r) ≤ Q. 1 ≤ i ≤ Nb. (10). h=1. qi,h (r) ≤ qi ,. i ∈ Dh ,. (11). and III. JPRA A LGORITHM The problem of the joint PA and RA for multirate SHOs is here defined as a constrained combinatorial optimization problem with an objective to maximize the total throughput of multirate SHOs, under the constraints of that the total power allocated to SHO users in each cell should be bounded by a maximum value, the power allocated to each SHO user from a cell of its active set is limited by an upper bound, and the. γh (r) ≥ γ ∗ (r),. ∀ h,. (12). i is the maximum value of total allocation power of where Q cell i for SHOs, qi is the upper bound of the PA to SHO user h from cell i, and Nb is the numbers of BTSs in the system. We propose a novel JPRA algorithm to efficiently solve the above optimization problem. The JPRA algorithm is mainly.
(5) 1392. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 56, NO. 3, MAY 2007. composed of the LPPA scheme, which determines all possible qi,h (r) under the constraints of (11) and (12), and the ECRA method, which searches for R∗ of (9) under the constraint of (10). In this joint PA and RA for SHO users, power balancing can be accomplished among cells. A. LPPA Scheme The LPPA scheme is an iterative algorithm to determine the transmission power required for SHO user h, qh (r), and how much amount of qh (r) would be from cell i in its active set Dh , qi,h (r), under the constraint of the maximum link power, qi , i ∈ Dh , as stated in (11). The resultant qi,h (r) is proportional to the link quality between the BTS i ∈ Dh and the SHO user h [10]. That is, qi,h (r) = qh (r) × i,h , where i,h is the weighting factor of the required transmission power for the link between BTS i and user h. We set i,h by Li,h . i,h = Li,h. (13). i∈Dh. Basically, the link proportional strategy for PA is to let the active link with better link quality contribute more power than those with weaker link qualities. If the required transmission power of one link violates the constraint of the maximum link power, LPPA will compensate the required power through other links by using an iterative method to redistribute qh (r) to all serving BTSs to satisfy the required signal quality. The iteration is to try to accomplish the power balance among mixed-size cells. Also, the resultant signal quality of user h with service rate r should not be less than its required signal quality γh∗ (r), as stated in (12); otherwise, the qh (r) should be adjusted by a tuning factor φh , which is given by φh =. γh∗ (r) . γh (r). Step 1: [Exam the SHO feasibility] • Set φh = 1.0. • Allocate maximum link power qi for each active links i. • Calculate received signal quality γh (r) based on (2), (3). • IF γh (r) > γ ∗ (r), THEN goto Step 2. ELSE IF γh (r) = γ ∗ (r), THEN set qi,h (r) = qi , i ∈ Dh , and goto Step 6. ELSE IF r = rv or r = rd1 , THEN SHO user h is forced to terminate such that qi,h (r) = 0, i ∈ Dh , DONE. ELSE goto Step 6. Step 2: [Initialize power settings] • Initialize required transmission power qh (r) for SHO user h by qh (r) = i∈Dh qi . Step 3: [Calculate weighting factor i,h ] • Obtain weighting factor i,h for the transmission power from BTS i in Dh to SHO user h, based on Li,h by (13). Step 4: [Calculate allocation power qi,h (r)] • Determine the power that BTS i in Dh allocates to SHO user h, qi,h (r), by qi,h (r) = min {qh (r) × i,h , qi } ,. ∀ i ∈ Dh .. (15). Step 5: [Compute received Eb /No and tuning factor φh ] • Compute the corresponding γh (r) based on (2) and (3), and set tuning factor φh based on (14). Step 6: [Check Stop Criterion] • IF φh = 1.0, THEN let qh (r) = φh × qh (r) and goto Step 2. ELSE IF r ∈ rd and r = rdM , THEN m = m + 1, r = rdm and goto Step 1. ELSE DONE.. (14). Besides, it is noteworthy that due to the constraint of the maximum link power, there exists a forced termination situation for SHO whenever SHO users cannot obtain the required signal quality even though all active links are allocated with maximum link power. If the SHO user is forced to be terminated, qi,h (r) of each link i in active set Dh is reset to zero. We have proven that the LPPA scheme is convergent in [10]. Here, the LPPA scheme will calculate all PA combinations of qi,h (r), 1 ≤ h ≤ Nv + Nd , 1 ≤ i ≤ Nb and r ∈ {rv , rd1 , . . . , rdM }, for the joint PA and RA problem described in (9)–(12). Here, only the determination of qi,h (r) for an arbitrary SHO user h with rate r ∈ {rd1 , . . . , rdM } is stated. In the following, we brief the LPPA scheme, and the details can be found in [11]. [The LPPA Scheme] Step 0: [Initialize service rate r for SHO user h] • IF SHO user h is with voice service, THEN set r = rv . ELSE Set r = rdm and m = 1 for SHO user h is with data service.. B. ECRA Method The ECRA method performs optimal RA for multirate SHO users based on the formulation in (9)–(12). It is found that even after we have reduced the complexity of the global optimization problem by taking care of joint PA and RA only for multirate SHO users, the computation time would still be a major concern for practical applications, particularly when there is a larger number of multirate SHO users being managed. Recall that there are Nd (Nv ) SHO data (voice) users and M kinds of data service rate in the system. The LPPA scheme has determined the qi,h (r), i ∈ Dh for r = {rd1 , . . . , rdM } if HO data user and for r = rv if HO voice user, 1 ≤ h ≤ Nd + Nv . The searching complexity is larger than (M + 1)Nd by using exhaustive method, in which 1 means a zero service rate for suspending transmission. If Nd is 10 and M is 4, there are nearly 107 searching complexity. This is far beyond a reasonable computation time for feasible application. In this paper, an evolutionary computing algorithm [18], which is a promising intelligent technique to effectively search a global optimal solution, is adopted..
(6) LIAO et al.: JPRA ALGORITHM FOR MULTIRATE SOFT HANDOFFS IN MIXED-SIZE WCDMA CELLULAR SYSTEMS. The evolutionary computing technique is a more advanced genetic algorithm; it uses stochastic searches to solve difficult optimization problems in real world through simulating natural genetic processes of living organisms, including selection, mutation, and crossover, to come up with better populations with better chromosomes generation by generation. It represents the service rate of each HO user as a chromosome, and a collection of chromosomes forms a population. Each population is regarded as the possible solution for the optimal RA of all HO data users. For M kinds of data service rates, each rate rd is encoded into log2 (M + 1) binary digits, denoted by chromosome x, and the decoder function for x is denoted by s(x). Thus, for SHO data user h with service rate rd , its corresponding allocation power is qh (s(xh )), in which the allocation power from active link i by the LPPA scheme is qi,h (s(xh )). Recall that rv is the service rate for SHO voice users and qi,h (rv ) is the corresponding allocation power from active link i to SHO user h. The ECRA method is to find an optimal RA vector (decision vector) for Nd multirate SHO users, x∗ = [x∗1 , x∗2 , . . . , x∗Nd ], by maximizing the objective function O(x), which is defined to be the total throughput of SHO data users. We restate the optimization problem defined in (9) and (10) as ∗. x = arg max O(x) ≡ x. Nd . s(xh ). (16). h=1. subject to constraints Nv . Nd . qi,h (rv ) +. h=1. i , qi,h (s(xh )) ≤ Q. 1 ≤ i ≤ Nb .. h=1. (17). Note that each x reveals different degree of the power balance among cells, and x∗ in (16) is corresponding to R∗ = [s(x∗1 ), . . . , s(x∗Nd )] in (9). Because of the constraint, some decision vectors may be out of the feasible domain. A violation function is used to rank the violation degree of the decision vector [18]. The value of the violation function indicates how far the solution deviates from the feasible region. This constrained violation function is defined as Nb. i=1. Ψ(x) = +. Mi. Nd. h=1. . Nv h=1. qi,h (rv ) +. Nd. h=1. qi,h (s(xh )). 2. 2Nb Hh [γh (s(xh ))]2 2Nd. (18). where Mi and Hh are the Heaveside operators [18], in which Mi [·] = 1 {Hh [·] = 1} whenever the constraint in (17) is violated, and Mi [·] = 0 {Hh [·] = 0} otherwise. In the following, the ECRA method is described. Noticeably, the allocation powers for SHOs are corresponding to the ones obtained by the LPPA scheme. [The ECRA method] Step 1: [Initialize]. 1393. • Set crossover rate pc , mutation rate pu , and maximum number of generations T . • Initialize generation t = 1, optimal objective value O∗ = 0, and optimal decision vector x∗ = 0 (zero pattern). • Generate K populations that are randomly selected decision vectors xk = [xk1 , . . . , xkNd ], 1 ≤ k ≤ K. Step 2: [Execute constraint tournament selection to form x k ] • Choose K tournament pairs randomly among xk , and calculate Ψ(xk ) (18) for each competitive pair. • Replace xk with winner x k which has minimum Ψ(xk ). Step 3: [Execute variable point crossover to form x. k ] • Choose K/2 crossover pairs from adjacent population x k and x k+1 , where k is odd. • Generate a random number c in [0, 1] for each chromosome in each crossover pair. • For the chromosome with c < pc , generate the crossover point randomly in [1, log2 (M + 1)] and make the crossover operation within it. Step 4: [Execute uniform mutation to form x. k] • Generate a random number u in [0, 1] for every bit in each population x. k. • Mutate any bit with u < pu . Step 5: [Update O∗ and x∗ ]. • Find feasible populations {x. k } with Ψ(xk ) = 0 ∗. ∗ • Set O = max{O(xk )} and x = argxk max{O(x. k )}. Step 6: [Check the stop criterion] • IF t < T , THEN Set t = t + 1, and Goto Step 2. ELSE DONE.. IV. S IMULATION R ESULTS AND D ISCUSSION A. Simulation Model Consider an example of mixed-size WCDMA cellular system with 12 wrapped around squared cells, including four microcells in the central congested region and eight macrocells in the neighboring cells [11]. The radii of macrocell and microcell are RM = 1 km and Rµ = 1/2 km, respectively; thus, the cell radius ratio between microcell and macrocell is ρ = Rµ /RM = 1/2. The antenna height of BTS in macrocell (microcell) are 20 (10) m, and the antenna height of mobile station is 1.5 m. For the propagation channel model, both pathloss and longterm shadowing are taken into account, in which two slope pathloss exponents are 2 and 4 dB, and the standard deviations of two slope shadowing are 4 and 8 dB [24]. Note that we also consider the cases of ρ = 1 and ρ = 1/3 in Fig. 8 to show the performance results versus different cell size ratios in mixedsize cell architecture. Furthermore, assume the bandwidth W is 4.096 MHz, and each cell utilizes the same frequency, in which the orthogonality factor is 0.5. For the power budget design, Pi for macrocell (microcell) i is 20 (10) watts, and qi of the macrocell (microcell) is 1 (0.5) watts. Here, η is 2 dB, and the maximum active set size is 3. In simulations, two cases of which there are without and with MEs during the active set selection are concerned. For the case with MEs, the received signal strength of each user is added an error signal that is Gaussian distributed random variable with zero mean and 1.5 dB standard deviation..
(7) 1394. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 56, NO. 3, MAY 2007. TABLE I SERVICE CLASSES. Users are assumed to be uniformly distributed in each cell, and each user moves in a constant speed of 36 km/h. The probability of moving direction change for users is 0.2, and the range of each direction angle change is among ±45◦ [25]. By mobility, the correlated shadowing effect is based on Gudmundson model [26], in which the decorrelation length is equal to 20 m in a vehicular environment. Assume the shadowing factor will not be varied when the moving distance is less than 4 m and there are five averaging windows in each snapshot. For 36 km/h mobility speed, the correlated shadowing duration is 400 ms. Also, assume the allowable number of iteration for system operation is 40, and each iteration takes 1 frame time (10 ms). Performance measurements are averaged from 2000 independent instances of user location and shadowing, and each snapshot has five correlated instances. For parameters of the ECRA method in the JPRA algorithm, the supportable service classes and the corresponding codes for the ECRA method are listed in Table I. The population size K is 100, the crossover rate pc is 0.5, the mutation rate pm is 0.05, and the stop generation T is 20. In simulations, each cell has the same numbers of voice and data users; thus, the central four microcells form a highly traffic congested region. Assume the number of voice users Nv is 30 and that of data users Nd ranges from 3 to 12 in each cell. B. Results and Discussion There are two essential performance measures investigated. One is the HO forced termination probability, which indicates the service continuity and the effectiveness of the cell’s service coverage. It is evaluated by counting the proportion of SHO users that are terminated by the system due to insufficient power resource for SHOs temporarily. The other is the total throughput of the system, which is obtained by summing all allocated transmission rates of users, which represents the system capacity. Because of limited downlink power resources of one BTS, under various nonuniform traffic load situations, there exist tradeoffs between the cell’s service coverage and the system capacity using different joint PA and RA schemes. The proposed JPRA algorithm is compared with SSDT [8] and LPPA schemes with the best effort RA. The best effort RA is to assign HO users maximum allowable transmission rate which exists feasible PA solutions to satisfy the users’ required quality of services. In order to achieve fair comparison for all schemes, the total power constraint of SHOs is confined to 0.3 times maximum transmission power of each BTS [27]. In simulations, we take the best effort rate allocation as the benchmark for comparisons, and denote SSDT and LPPA with the best effort RA by SSDT and LPPA, respectively.. Fig. 3. Averaged forced termination probability of SHO without ME and with 1.5 dB ME.. Fig. 3 shows the average forced termination probability of SHO under different traffic loads. It can be seen that the HO forced termination probability of JPRA is much less than those of SSDT and LPPA schemes by over 300% and around 200%, respectively. It can be inferred that the significant gain of JPRA over LPPA comes from the RA for SHO by the ECRA method, and the gain of LPPA over SSDT comes from the PA strategy for SHO [10], [11]. The reasons are that the allowable transmission rates are highly correlated to the feasible PA solutions; RA for each SHO user directly affects the management of power resource in at least two BTSs; thus, an optimal RA for SHO users obtained by the ECRA method will further enhance the effect of power balance among cells. Also, it can be seen that, when the traffic load is light, LPPA achieves less average forced termination probability of SHO than SSDT. When the traffic load is heavy, LPPA performs worse than SSDT. The reason is that higher interferences are induced by the multisite transmission mechanism than by the single-site transmission mechanism for HOs. The finite total power resource of the BTS may be insufficient to support more HO users at heavy load situations. Moreover, consider the case with ME during the active set selection. It is observed that MEs incur higher HO forced termination probabilities for all schemes because BTSs in active set waste more power on multirate HO users. Also, the superiority of JPRA over SSDT is increased up to around 500%. This is because the outstanding power balance characteristic of JPRA can release the effect of MEs. If the HO forced termination probability is regraded as the performance index of the cell’s service coverage, in which smaller HO forced termination probability means better cell coverage, JPRA achieves better cell’s service coverage than SSDT and LPPA. JPRA indeed provides the superb power balance characteristic and optimal RA for multirate SHOs; thus, it cannot only provide better service continuity performance but also possess the capability of the resistance to MEs. Fig. 4 shows the total HO throughput versus different number of data users. It is found that both LPPA and JPRA have higher throughput than SSDT because of the multisite transmission.
(8) LIAO et al.: JPRA ALGORITHM FOR MULTIRATE SOFT HANDOFFS IN MIXED-SIZE WCDMA CELLULAR SYSTEMS. Fig. 4. Total handoff throughput versus the number of data users per cell without ME and with 1.5 dB ME.. Fig. 5. Average call dropping probability versus the number of data users per cell without ME and with 1.5 dB ME.. mechanism, and JPRA achieves capacity gain over LPPA as in the aspect of the HO force termination probability shown in Fig. 3. The reason is that there exists a tradeoff between coverage and capacity for JPRA and LPPA schemes due to the total power constraint of SHOs. The ECRA method in the JPRA reduces the average transmission rate of multirate SHOs so as to accomplish better cell coverage while LPPA leads to more terminated HO users and, thus, higher HO throughput than JPRA. The similar results could be observed in the case of MEs. But, since higher HO forced termination probability in this case results in more power left for survived HO users to transmit with higher average transmission rates, the total HO throughput in the case of MEs is higher than those in the case of error free. In addition, we can see that the ECRA method in JPRA plays an important role to reduce performance degradation by MEs. This is because the ECRA method allocates optimal transmission rates for multirate SHOs to further balance power loads among cells. Therefore, JPRA successfully enhances cell’s service coverage and HO throughput.. 1395. Fig. 6. Total system throughput gain versus the number of data users per cell without ME and with 1.5 dB ME.. Fig. 5 shows the average call dropping probability versus the number of data users. The call will be dropped when there is no feasible PA solution to support the user with its required service qualities for a period of time. When the required average call dropping probability is 0.01, JPRA (LPPA) can improve the average call dropping probability of all users by 71% (0%) and 200% (25%) without and with MEs, compared to SSDT. It is mainly because JPRA owns a better power balance characteristic via multirate SHO resulting from LPPA and ECRA. This optimal radio resource management can achieve an appropriate assignment among BTSs in the active set, preventing one BTS from using too much power resources to serve multirate SHOs. Thus, there would be more power resource preserved to serve NHO users. Fig. 6 shows the total system throughput gain, referred to SSDT. We can see that JPRA (LPPA) can achieve a higher total system throughput than SSDT by 5% (3%) and 8% (4%) in the case of ME free and 1.5 dB MEs, respectively. The reasons are that the power balance characteristic of JPRA and LPPA schemes can aid the system to balance the power load more among cells, and the multisite transmission can support higher transmission rates for SHO users. JPRA indeed makes great improvements in cell’s service coverage and system capacity because it can optimally allocate resource for multirate SHOs. However, it is found that when the number of data users gets larger, the gain gets worse. This is because the multisite transmission induces larger interference that results in the degradation of the system throughput. In the meantime, on the perspective of the user satisfaction, voice and data users should have different service requirements. Denote the call dropping probabilities of voice (data) users as Pv (Pd ). Also, the summation of allocated (requested) transmission rates of data users is represented by Rd (R∗d ). Two USIs for voice and data services, denoted by USIv and USId , respectively, are defined as. USIv = (P∗v − Pv ) /P∗v USId = κd × Rd /R∗d + (1 − κd ) × (P∗d − Pd ) /P∗d. (19).
(9) 1396. Fig. 7. USI versus the number of data users per cell for (a) USI of voice users (USIv ) and (b) USI of data users (USId ), respectively.. Fig. 8. Total system throughput for the SSDT, LPPA, and JPRA schemes under different cellular architecture in cell radius ratio ρ = 1, 1/2, and 1/3, with and without ME.. where κd is the weighting factor for the total throughput and call dropping probability of data users, and Pv∗ (Pd∗ ) is the system requirement for voice (data) call dropping probability. Fig. 7(a) and (b) shows USIv and USId versus the number of data users, respectively, where both P∗v and P∗d are set as 0.05, and assume κd is 0.7 because data users usually are satisfied with higher transmission rates and can tolerate longer transmission delay. It is shown that the proposed JPRA scheme can provide better USId , whereas it still could maintain good USIv even when there exist MEs. Fig. 8(a) and (b) shows the total system throughput for the SSDT, LPPA, and JPRA schemes under cell radius ratio ρ = 1, 1/2, and 1/3 with and without MEs, respectively. It is found that when the number of data users is smaller than. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 56, NO. 3, MAY 2007. nine, JPRA always achieves the highest total throughput for any cell radius ratio. In particular, for the mixed-size cellular system with smaller cell radio ratio, JPRA can achieve the capacity gain larger than the others because of its outstanding power balance characteristic. In the case of MEs, we can see that, compared to without ME case, JPRA has much higher gain of the system throughput than SSDT and LPPA. When the cell radius ratio is 1/3 and the number of data users is getting larger, the total system throughput of the multisite transmission schemes, LPPA and JPRA, is degraded severely. This is because multisite transmission schemes for HO users may cause large interference in the highly congested mixed-size WCDMA cellular system. Also, when the system is overloaded and exceeds predetermined power budget, in order to release some power load and preserve users’ required signal quality as much as possible, JPRA adjusts the transmission rates of the SHO data users aggressively. But, LPPA cannot respond to the variation of the interference promptly due to the lack of rate adaptation function. The LPPA has decrement rate of system throughput less than JPRA. On the other hand, SSDT is very suitable for applying to the highly congested environment with small cell radius ratio, 1/3 in this case, because the singlesite transmission scheme induces less interference than other multisite transmission ones. However, based on the previous results from Figs. 3–7, SSDT gets the worst performance of USId for cell radius ratio of 1/2, in which USId is related to not only throughput but also call dropping probability for data users. In the case of cell radius of 1/3, although JPRA could not achieve better system throughput when the traffic load is highly congested, compared to SSDT, it can still achieve better USId because of smaller call dropping probability. As for the searching complexity issue, it is defined as the number of operations for all searching patterns needed to come up with an optimal rate allocation for all multirate SHO users. In the case of the exhaustive search, the searching complexity is (M + 1)Nd . In the case of the ECRA method, the searching complexity is given as log2 (M + 1) × K × T × Nd , upper bounded by the multiplication of the length of populations, number of SHO data users Nd , population size K, and number of generation T . Note that M is the number of supportable service rates in the system. It can be found that when the number of multirate data users is larger than seven, the searching complexity of the exhaustive search grows exponentially. The ECRA method can efficiently reduce the searching complexity, compared to the exhaustive search. V. C ONCLUSION In this paper, a JPRA algorithm is proposed to deal with multirate SHOs in WCDMA mixed-size cellular systems. It contains the LPPA scheme and the ECRA method. Compared to SSDT and LPPA schemes with best effort based RAs, simulation results show that JPRA accomplishes superior power balance among cells so that the JPRA algorithm can achieve better cell’s service coverage and higher system capacity. Also, JPRA can offer greater user satisfaction for voice and data users. Furthermore, JPRA is less sensitive to MEs in the active set selection than SSDT with best effort rate allocation..
(10) LIAO et al.: JPRA ALGORITHM FOR MULTIRATE SOFT HANDOFFS IN MIXED-SIZE WCDMA CELLULAR SYSTEMS. It is noteworthy that the aforementioned advantages of JPRA are more conspicuous in WCDMA mixed-size cellular systems with smaller cell radius ratio between microcell and macrocells. Also, the evolutionary computing algorithm is applied for the first time to solve the downlink resource optimization problem for SHO in this paper; the JPRA algorithm can efficiently reduce the complexity problem of downlink resource management and make real system implementation feasible. ACKNOWLEDGMENT The authors would like to thank the anonymous reviewers for their very thoughtful and valuable comments, making the presentation of this paper better. R EFERENCES [1] S. L. Kim, Z. Rosberg, and J. Zander, “Combined power control and transmission rate selection in cellular networks,” in Proc. IEEE VTC—Fall, Amsterdan, The Netherlands, Sep. 1999, pp. 1653–1657. [2] C. W. Sung and W. S. Wong, “Power control and rate management for wireless multimedia CDMA systems,” IEEE Trans. Commun., vol. 49, no. 7, pp. 1215–1226, Jul. 2001. [3] M. Soleimanipour, W. Zhuang, and G. H. Freeman, “Optimal resource management in wireless multimedia wideband CDMA systems,” IEEE Trans. Mobile Comput., vol. 1, no. 2, pp. 143–160, Apr./Jun. 2002. [4] D. Kim, “Rate-regulated power control for supporting flexible transmission in future CDMA mobile networks,” IEEE J. Sel. Areas Commun., vol. 17, no. 5, pp. 968–977, May 1999. [5] Y. W. Kim, D. K. Kim, J. H. Kim, S. M. Shin, and D. K. Sung, “Radio resource management in multiple-chip-rate DS/CDMA systems supporting multiclass services,” IEEE Trans. Veh. Technol., vol. 50, no. 3, pp. 723–736, May 2001. [6] D. I. Kim, E. Hossain, and V. K. Bhargava, “Downlink joint rate and power allocation in cellular multirate WCDMA systems,” IEEE Trans. Wireless Commun., vol. 2, no. 1, pp. 69–80, Jan. 2003. [7] S. Kahn, M. K. Gurcan, and O. O. Oyefuga, “Downlink throughput optimization for wideband CDMA systems,” IEEE Commun. Lett., vol. 7, no. 5, pp. 251–253, May 2003. [8] H. Furukawa, K. Hamabe, and A. Ushirokawa, “SSDT—Site selection diversity transmission power control for CDMA forward link,” IEEE J. Sel. Areas Commun., vol. 18, no. 8, pp. 1546–1554, Aug. 2000. [9] 3GPP Technical Specification 25.922, Radio resource management strategies, Dec. 1999. [10] C. Y. Liao, L. C. Wang, and C. J Chang, “Power allocation mechanisms for downlink handoff in the CDMA system with heterogeneous cell structures,” ACM/Kluwer WINET, vol. 11, no. 5, pp. 593–605, Sep. 2005. [11] C. Y. Liao, L. C. Wang, and C. J Chang, “A joint power and rate assignment algorithm in mixed-size WCDMA cellular systems,” in Proc. ACM IWCMC, Vancouver, BC, Canada, Jul. 2006, pp. 833–838. [12] A. Morimoto, K. Higuchi, and M. Sawahashi, “Site independent diversity transmit power control for inter-cell site diversity in W-CDMA forward link,” in Proc. IEEE VTC—Fall, Atlantic City, NJ, Oct. 2001, pp. 645–649. [13] D. Staehle, K. Leibnitz, and K. Heck, “Effects of soft handover on the UMTS downlink performance,” in Proc. IEEE VTC—Fall, Vancouver, BC, Canada, Sep. 2002, pp. 960–964. [14] F. Blaise, L. Elicegui, F. Goeusse, and G. Vivier, “Power control algorithms for soft handoff users in UMTS,” in Proc. IEEE VTC—Fall, Vancouver, BC, Canada, Sep. 2002, pp. 1110–1114. [15] H. G. Jeon, S. M. Shin, T. Hwang, and C. E. Kang, “Reverse link capacity analysis of a CDMA cellular system with mixed cell sizes,” IEEE Trans. Veh. Technol., vol. 49, no. 6, pp. 2158–2163, Nov. 2000. [16] S. Kishore, L. J. Greenstein, H. V. Poor, and S. C. Schwartz, “Uplink user capacity in a CDMA macrocell with a hotspot microcell: Exact and approximate analyses,” IEEE Trans. Wireless Commun., vol. 2, no. 2, pp. 364–374, Mar. 2003. [17] S. Kishore, L. J. Greenstein, H. V. Poor, and S. C. Schwartz, “Downlink user capacity in a CDMA macrocell with a hotspot microcell,” in Proc. IEEE GLOBECOM, San Francisco, CA, Dec. 2003, pp. 1573–1577. [18] A. Osyczka, Evolutionary Algorithms for Single and Multicriteria Design Optimization. Heidelberg, Germany: Physica-Verlag, 2002.. 1397. [19] S. A. Grandhi, J. Zander, and R. D. Yates, “Constrained power control,” Wireless Pers. Commun., vol. 1, no. 4, pp. 257–270, 1995. [20] D. Kim, “A simple algorithm for adjusting cell-site transmitter power in CDMA cellular systems,” IEEE Trans. Veh. Technol., vol. 48, no. 4, pp. 1092–1098, Jul. 1999. [21] M. Andersin, Z. Rosberg, and J. Zander, “Gradual removals in cellular PCS with constrained power control and noise,” ACM/Baltzer Wireless Netw. J., vol. 2, no. 1, pp. 27–43, 1996. [22] F. Berggren, R. Jantti, and S. L. Kim, “A generalized algorithm for constrained power control with capability of temporary removal,” IEEE Trans. Veh. Technol., vol. 50, no. 6, pp. 1604–1612, Nov. 2001. [23] S. L. Kim, “Optimization approach to prioritized transmitter removal in a multiservice cellular PCS,” in Proc. IEEE PIMRC, Boston, MA, Sep. 1998, pp. 1565–1569. [24] V. Erceg, S. Ghassemzadeh, M. Taylor, D. Li, and D. L. Schilling, “Urban/suburban out-of-sight propagation modeling,” IEEE Commun. Mag., vol. 30, no. 6, pp. 56–61, Jun. 1992. [25] Universal Mobile Telecommunication System (UMTS), Selection procedures for the choice of radio transmission technologies of the UMTS, pp. 54–55, 1998. UMTS 30.03, version 3.2.0, TR 101 112. [26] M. Gudmundson, “Correlation model for shadow fading in mobile radio systems,” Electron. Lett., vol. 27, no. 23, pp. 2145–2146, Nov. 1991. [27] A. J. Viterbi, CDMA: Principles of Spread Spectrum Communication. Reading, MA: Addison-Wesley, Jun. 1995, pp. 218–224.. Ching-Yu Liao (S’02–M’05) received the Ph.D. degree in communication engineering from National Chiao Tung University, Hsinchu, Taiwan, R.O.C., in 2004. From January 2004 to May 2005, she was a Visiting Scholar with the University of British Columbia, Vancouver, Canada. She is currently a Senior Research Scientist at Telcordia Applied Research Center Taiwan Company (TARC-TW), Taipei, Taiwan. Also, she has been an Adjunct Assistant Professor with National Taiwan University of Science and Technology (NTUST), Taipei, Taiwan, since January 2007. Her research interests include handoff techniques, radio resource management, heterogeneous wireless networks, cross-layer design, stochastic optimization, etc.. Chung-Ju Chang (S’84–M’85–SM’94–F’06) was born in Taiwan, R.O.C., in August 1950. He received the B.E. and M.E. degrees in electronics engineering from National Chiao Tung University (NCTU), Hsinchu, Taiwan, in 1972 and 1976, respectively, and the Ph.D. degree in electrical engineering from National Taiwan University (NTU), Taipei, Taiwan, in 1985. From 1976 to 1988, he was with Telecommunication Laboratories, Directorate General of Telecommunications, Ministry of Communications, Taiwan, as a Design Engineer, Supervisor, Project Manager, and then Division Director. In the meantime, he also acted as a Science and Technical Advisor for the Minister of the Ministry of Communications from 1987 to 1989. In 1988, he was with the Faculty of the Department of Communication Engineering, College of Electrical Engineering and Computer Science, NCTU, as an Associate Professor. He has been a Professor since 1993. He was a Director of the Institute of Communication Engineering from August 1993 to July 1995, a Chairman of the Department of Communication Engineering from August 1999 to July 2001, and the Dean of the Research and Development Office from August 2002 to July 2004. Also, he was an Advisor for the Ministry of Education to promote the education of communication science and technologies for colleges and universities in Taiwan during 1995–1999. His research interests include performance evaluation, wireless communication networks, and broadband networks. Dr. Chang is a member of the Chinese Institute of Engineers (CIE). He is acting as a Committee Member of the Telecommunication Deliberate Body, Taiwan. He serves as Editor for IEEE Communications Magazine and Associate Editor for IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY..
(11) 1398. Li-Chun Wang (M’96–SM’05) received the B.S. degree from National Chiao Tung University (NCTU), Hsinchu, Taiwan, R.O.C., in 1986, the M.S. degree from National Taiwan University, Taipei, Taiwan, in 1988, and the Ms.Sci. and Ph.D. degrees from Georgia Institute of Technology, Atlanta, in 1995 and 1996, respectively, all in electrical engineering. From 1990 to 1992, he was with the Telecommunications Laboratories of the Ministry of Transportations and Communications in Taiwan (currently the Telecom Labs of Chunghwa Telecom Company). In 1995, he was affiliated with Bell Northern Research of Northern Telecom, Inc., Richardson, TX. From 1996 to 2000, he was with AT&T Laboratories, where he was a Senior Technical Staff Member in the Wireless Communications Research Department. Since August 2000, he has been in the Department of Communication Engineering, NCTU, as an Associate Professor and has been promoted to a Full Professor in August 2005. His current research interests are in the areas of cellular architectures, radio network resource management, cross-layer optimization, and cooperation wireless communications networks. He is the holder of three U.S. patents. Dr. Wang was a corecipient of the 1997 IEEE Jack Neubauer Best Paper Award from the IEEE Vehicular Technology Society. He is an Associate Editor for the IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS .. IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 56, NO. 3, MAY 2007. Yih-Shen Chen (S’97–A’97–S’01–M’04) was born in Taiwan, R.O.C., in September 1973. He received the B.E., M.E., and Ph.D. degrees in communication engineering from National Chiao Tung University, Hsinchu, Taiwan, in 1995, 1997, and 2004, respectively. In 2005, he joined a 3G project in Sunplus Technology Company, Ltd., Hsinchu. His research interests include performance analysis, protocol design, and personal communication service (PCS) networks..
(12)
相關文件
Then, we tested the influence of θ for the rate of convergence of Algorithm 4.1, by using this algorithm with α = 15 and four different θ to solve a test ex- ample generated as
In this paper, we extended the entropy-like proximal algo- rithm proposed by Eggermont [12] for convex programming subject to nonnegative constraints and proposed a class of
In summary, the main contribution of this paper is to propose a new family of smoothing functions and correct a flaw in an algorithm studied in [13], which is used to guarantee
For the proposed algorithm, we establish a global convergence estimate in terms of the objective value, and moreover present a dual application to the standard SCLP, which leads to
In this paper, we develop a novel volumetric stretch energy minimization algorithm for volume-preserving parameterizations of simply connected 3-manifolds with a single boundary
In this paper, a novel subspace projection technique, called Basis-emphasized Non-negative Matrix Factorization with Wavelet Transform (wBNMF), is proposed to represent
In this chapter, a dynamic voltage communication scheduling technique (DVC) is proposed to provide efficient schedules and better power consumption for GEN_BLOCK
This bioinformatic machine is a PC cluster structure using special hardware to accelerate dynamic programming, genetic algorithm and data mining algorithm.. In this machine,
