Priority-based Resource Allocation to Guarantee ...

Priority-based Resource Allocation to Guarantee Handover and Mitigate Interference for OFDMA System Lexi Xu, Yue Chen School of Electronic Engineering and Computer Science Queen Mary, University of London London, UK Abstract—Mitigating Inter-cell Interference (ICI) and ensuring seamless high-quality communication are two challenging issues for OFDMA systems. Cell-level coordinated resource allocation and handover (HO) are the two key technologies for achieving these goals. They have been investigated intensively, however, mainly separately. In this paper, a novel combined Handover Guarantee and Interference Mitigation (HGIM) cell-level resource allocation scheme is proposed. HGIM defines the Handover User Set (HUS) and grants higher allocation priority to handover users. Other active users are prioritized based on a unified cell division model which divides a cell into different ICI sensitive areas. Meanwhile, HGIM defines a Sub-carrier Preferred List (SPL) to optimize allocation. Simulation results show that HGIM achieves greater ICI mitigation and improved handover performance compared with the conventional soft frequency reuse scheme. Keywords—OFDMA; Inter-cell interference (ICI); Handover; Interference mitigation;

I.

INTRODUCTION

OFDMA (Orthogonal Frequency Division Multiple Access) is considered as the leading multiple access technique for future wireless systems, such as LTE, IMT-Advanced, as it provides high spectral efficiency and robust performance in a fading environment [1]. However, multi-cell OFDMA systems are sensitive to Inter-cell Interference (ICI). ICI has a negative impact on the communication quality especially for cell edge users. Inevitably, ICI affects the Handover (HO) performance as well because handover users are cell edge users. This imposes a challenge for future systems on coordinating celllevel resource allocation in order to achieve ICI mitigation and better handover performance under the bandwidth restriction. In recent years, cell-level resource allocation has become a hot research topic. Most work is based on the Soft Frequency Reuse (SFR) concept proposed in [2], which assigns a third of all sub-carriers to the cell edge, with a different third for neighbouring cells, so reducing ICI because of the frequency reuse factor of 3 for the cell edge. Semi-dynamic coordination has been proposed and investigated in [3]-[5]: this is more flexible than SFR and has better cell edge performance. Some other schemes, such as soft fractional reuse and other partial reuse schemes [6][7] based on the SFR concept have also been proposed, but the complex cell edge separation causes high user area-selection mistakes. For example, in [6], the cell edge part is separated into 12 complex areas with different resource units. Most existing cell-level resource allocation schemes do

not differentiate handover users from other active users with the performance evaluation focusing on the cell edge in terms of throughput and fairness improvement. Meanwhile, there has been much research on handover for OFDMA systems, such as the Received Signal Strength (RSS) based hard handover for LTE [8], the fast handover for realtime service in [9], the soft handover [10], and the semi-soft handover proposed in [11]. Most of the work focuses on the handover performance without considering the ICI mitigation. In this paper, we propose a novel combined Handover Guarantee and Interference Mitigation (HGIM) cell-level resource allocation scheme for OFDMA systems. Handover users are identified and registered in the Handover User Set (HUS): they are granted with higher resource allocation priorities by mapping to the Sub-carrier Prefer List (SPL) before other active users. The conventional SFR scheme for cell-level interference coordination is also modified in this paper. A unified cell division principle is proposed that divides a cell into different ICI sensitivity areas associated with different Area Interference Mitigation Priority (AIMP). HGIM ensures handover users in HUS take preference in the allocation with good condition sub-carriers. For other active users, HGIM allocates preferred sub-carriers from the SPL to each area according to AIMP. This leads to a greater ICI mitigation compared to the conventional two-part, edge and inner, cell division approach. The rest of the paper is organized as follows. The HUS registration and the unified cell division model for interference coordination are presented in section 2. The novel HGIM scheme is proposed and analyzed in section 3 followed by the simulation results in section 4 and finally, the paper is concluded in section 5. II. HUS REGISTRATION & UNIFIED CELL DIVISION MODEL A. Handover Algorithm and HUS Registration As part of the mobility management functionality, handover is crucial for ensuring the seamless communication in cellular networks. In LTE, the Reference Signal Received Power (RSRP) based hard handover is chosen because of its advantages, such as reduced signalling load and less handover delay [8]. The handover process consists of 4 stages: measurements, processing, decision and execution, as shown in Fig. 1. Tm is handover measurement period and Tu is handover decision update period.

Fig.2 DL ICI analysis Fig.1 RSRP based handover process

In the measurement stage, a User Equipment (UE) measures the RSRP from its serving and neighbouring Base Stations (BS). The function of the layer 3 filtering is to filter out the effect of the fast fading and any imperfections in the layer 1 measurements. The handover criterion is checked in the decision stage. If the condition in (1) is satisfied for a preset time window, the time-to-trigger (TTT) window, the UE sends the measurement report to its serving BS. RSRPTC ≥ RSRPSC + Hm

(1)

RSRPTC and RSRPSC represent the RSRP from the target cell and the serving cell respectively. Hm is the handover margin, or handover hysteresis. In the execution stage, the UE is allocated with a sub-carrier from its new serving BS. The paper assumes ideal layer 3 filtering which means fast fading and measurement imperfections can be totally filtered out. Hence, the handover criterion can be expressed as (2) APGTC ≥ APGSC + Hm

(2)

APGTC and APGSC represent the Average Path Gain (APG) from the target cell and the serving cell respectively. Assuming UEs have the same average shadowing and antenna gain, the APG can be obtained in (3) by applying the UMTS30.03 path loss model APGSC (dB)=40lg(rSC)+30lg (f)+49 (3) Where rSC is the distance between the UE to its serving BS and f is the carrier frequency. All users satisfying (2) will be registered in the HUS and will be granted higher resource allocation priorities than other active users for two main reasons: i) Users in the HUS are cell edge users and require higher transmission power to meet the signal-to-interference ratio (SINR), this introducing more ICI to other co-channel users. ii) A handover failure affects user satisfaction more than a new call being blocked. In the simulation in this paper we assume a uniform distribution of users with 3 different speeds of movement and chooses a typical 3 dB handover margin[8] as the threshold for HUS registration. B. Interference Analysis & Unified Cell Division Model This paper focuses on the Downlink (DL) interference mitigation. For OFDMA systems, ICI is the main source of the interference as the intra-cell interference can be eliminated by filling the Guard Interval with the Cyclic Prefix of the OFDM symbol [12]. ICI can be analyzed using a generic two-cell scenario shown in Fig. 2. Interference analysis for multi-cell scenarios can be derived following the same principle.

In Fig. 2, user 0 is served by cell A and users 1 and 2 are served by cell B. ri represents the distance between the user i and its serving BS and di is the distance between the user i and the adjacent BS. In this particular scenario, r1>r2 and d1d1, from (4)(5)(6), it can be obtain that: ICI1>ICI2

(7)

SINR1< SINR2

(8)

After suffering severe interference, SINR of user 1, 2 could be less than the required threshold, SINRthr. To meet SINR1' = SINR2' = SINRthr , the expected received power of user j meets Pj' = SINRthr ( ICI j + N ) > Pj . Meanwhile, the transmitting

power

from

BS-B

to

user

j

is

PBS' − j =Pj' +40lg rj + 30lg f + 49 > PBS − j . So user 0 will suffer heavier ICI. From Fig. 2(b), suffix ‘j-0’ denotes parameters of user 0 after BS-B increases the transmitting power to user j. The ICI of user 0 meets:

ICI 'j −0 = PBS' − j − 40lg d0 − 30 lg f − 49

j ∈ [1, 2] (9)

From (4), decreasing the SINR of user 0 meets SINR'j −0 < SINR0 . Once SINR 'j −0 is less than SINRthr, BS-A will increase the transmitting power, then the ICI to the cell B will continue to increase. From (7) and (8), the expected received power has the relationship P1' > P2' , which indicates the expected received power of user 1 is higher than user 2. In

addition, due to r1>r2 and P1' > P2' , The BS transmitting power comparison meets:

PBS' −1 > PBS' − 2

(10)

The interference increment of user 0 from BS-B ΔICI 'j −0 can be calculated as:

ΔICI 'j −0 = ICI 'j −0 − ICI 0

j ∈[1, 2]

(11)

where ICI 0 = 0 (See Scene 1). From (7) (11), we can

Fig. 3 shows the unified cell division model used in this paper with a cell radius R=1000m. According to ICI distinction in (13), each cell is divided into 5 areas. Areas 1 to 5 are granted different resource allocation priorities. III.

THE NOVEL HGIM SCHEME

Based on the cell structure shown in Fig. 3, a novel combined Handover Guarantee and Interference Mitigation (HGIM) cell-level resource allocation scheme is proposed. Fig. 4 shows the flow chart of the HGIM scheme.

obtain:

ΔICI1'−0 > ΔICI 2' −0

(12)

(12) shows that user 1 induces severe ICI to user 0. From the above analysis, when users are assigned cochannel sub-carriers used in the adjacent cells, those who are located further from their serving BS will receive more ICI as shown in (7). Hence, greater transmitting power is needed from their serving BS than those users who are located nearer the cell centre (10). Meanwhile, it can be seen from (12) that users who are located further from their serving BS induce more ICI to the adjacent cell. Therefore, the area which is further from the BS should be preferably allocated with good channel condition sub-carriers for better interference mitigation performance. Based on the above findings, a unified cell division model is proposed which divides a cell into different interference sensitive areas. A Cell Interference Distinction Threshold (CIDT) is introduced, which specifies the minimum interference difference between adjacent divided areas of each cell. In this paper, CIDT is set to be 3dB. From (6), the ICI difference between the two adjacent areas can be expressed as:

ICI k − ICI ( k −1)

d 3R − rk −1 ≥CIDT (13) = 40 lg k −1 = 40 lg dk 3R − rk

Compared to the conventional two-part, edge and inner, cell division approach, the unified cell division model using CIDT has finer division granularity which leads to better ICI mitigation. On the other hand, this unified cell division model is much simpler than some of the soft fractional frequency reuse scheme as proposed in [6][7].

Fig.4 HGIM cell-level resource allocation scheme

The resource allocation can be summarized into three steps: A. Sub-carrier Priority Initialization of Each Cell Step 1: HGIM initializes the sub-carrier prioritization to guarantee the preferred resource of a particular cell is less likely to be assigned in the neighbouring cells. Assume there are L sub-carriers available in total. They are divided into three orthogonal sets, Q1, Q2, Q3. (j≠k; j,k=1,2,3) (14) Q j Qk = ∅

∩

Each set is composed of N sub-carrier denoted as they meet: Q1: A1, A2…AN Q2: B1, B2…BN Q3: C1, C2…CN N*3 ≤ L

Where Bi and Ci with the same suffix have the same priority in P1. Each cell assigns sub-carriers in its own orthogonal set, (Q1, Q2 and Q3 ). Take SPL-1 P1 as an example, cell-2 and cell-3 prefers to use Q2 and Q3 respectively, so are less likely to use Q1, namely these highest priority sub-carriers have best channel condition. Meanwhile, after A set is allocated, P1 prefers to assign low-priority sub-carriers of adjacent cell sub-carriers sets Q2 and Q3 with descending order of Bi and Ci. This SPL structure ensures that low priority sub-carriers are less likely to be used synchronously in adjacent cells, hence the ICI is mitigated. Fig.3 Unified cell division model for OFDMA systems

Fig.5 Sub-carrier Preferred List (SPL) structure Fig.6 HGIM resource allocation diagram

B. HUS and Area Interference Mitigation Prioritisation Step2: The HUS of each cell is defined as Hd 1, Hd 2 and Hd 3 respectively. All users satisfy (2) are registered to the relevant HUS and the allocation priority of HUS is set to be the highest: i PrHUS → ∞ ,1 ≤ i ≤ L

(15)

Applying the unified cell division model discussed in section 2, since the exterior circular area prefers to allocate resource with low ICI, the AIMP value of area 1 is the lowest and area 5 has the optimal priority shown as Table 1. Table 1 Detailed Area Interference Mitigation Priority (AIMP) Area N.O.(j) 1 area 2 area 3 area 4 area 5 area 271503698Area Range 0-271 >862 503 698 862 1-lowest 2 3 4 5(optimal) AIMP (gj)

C. Resource Allocation Step3: The actual sub-carrier allocation starts with users in HUS. Taking cell 1 for example, users in Hd 1 are allocated with the sub-carriers in P1 first. After allocation, these subcarriers will be deleted from the cell’s SPL. When resource allocation of HUS is completed, HGIM chooses the optimum match between the area ICI sensitivity and the SPL for other active users aiming at maximizing the ICI mitigation. To guarantee high interference sensitive areas are assigned with good condition sub-carriers, preferred subcarriers should be allocated to the with large AIMP value. So the allocation priority of area j on sub-carrier i of its cell SPL can be expressed as: i j

Pr = g j

1≤ j ≤ 5 , 1≤ i ≤ L

(16)

Where sub-carrier i of its cell SPL does not contain the allocated sub-carriers, it meets:

5

∑r

ji

After step 3, HUS takes the user-level handover process, and each area takes the user-level sub-carrier, power, bit allocation within the allocated sub-carriers. This paper only considers the cell-level sub-carrier allocation. IV.

A. Simulation Platform Configuration To verify the performance of the proposed HGIM scheme, a OFDMA system-level simulation platform was set up containing 36 cells with 12 three-cell clusters. Users are randomly distributed and the traffic arrival rate follows the Poisson distribution. Some other simulation parameters are shown in Table 2. Table 2 Simulation parameter Parameter Assumption Total Sub-carrier Number 192 Carrier Frequency 2GHz Cell Radius 1km Cell Division and AIMP 5 areas (See Table 1) User Speed 1,5,15 (m/s) Path Loss Model 40 lg (r) + 30 lg (f) + 49, r-km Fading Model Lognormal fading

The SFR scheme proposed in [2] is used as the reference scheme in this paper and its parameters are: cell edge covers 20% of cell radius with frequency reuse factor 3, and cell inner with frequency reuse factor 1. Other parameters are the same as HGIM. B. Simulation Results Fig.7 and Fig.8 compare the performance of the HGIM and the conventional SFR in terms of the handover dropping probability and the new call blocking probability respectively.

≤ 1 , rji ∈ [0,1] , and

0.08

j =1

The allocation

order for users except HUS can be written as:

(

W = arg max Prji j

j

)

0.07

1 ≤ j ≤ 5 ,1 ≤ i ≤ L

(17)

Sub-carrier i is assigned to the maximum Prji value area, and then the Prji value is deleted. The allocation process will recycle until all areas have been assigned with required resources or all sub-carriers have been allocated.

Handover Droping Probability

rji =1 denotes area j allocates sub-carrier i.

SIMULATION ANALYSIS

HGIM Scheme SFR Scheme

0.06 0.05 0.04 0.03 0.02 0.01 0 0.1

0.15

0.2 0.25 0.3 Poisson Arrival Rate(lamda) User/S

0.35

Fig.7 Handover dropping probability VS arrival rate

same network condition, the average ICI is lower and the average system throughput is higher with the HGIM scheme. This is also due to the optimum mapping between the ICI sensitivity areas and the sub-carriers on the SPL. Greater ICI mitigation leads to less overall interference in the system, which in return, increases the system throughput. V. CONCLUSIONS

0.16

New Call Blocking Probability

0.14


0.12 0.1 0.08 0.06 0.04 0.02 0 0.1

0.15


0.35

Fig.8 Call blocking probability VS arrival rate

A connection requesting handover is dropped when none of the remaining sub-carriers on the target BS SPL can meet the SINR requirement; a call is blocked when no sub-carrier on its serving BS SPL can meet the SINR requirement. It is clear the handover performance is improved with the HGIM scheme because handover users are identified in the HUS and granted higher resource allocation priority than other active users. Low ICI sub-carriers are allocated to the HUS that leads to great ICI mitigation The decreased call blocking probability is because users located in the exterior area, the area with high interference sensitivity, are assigned with better condition sub-carriers than users in the interior area, so they are more likely to get access to the serving BS.

In this paper, a novel combined Handover Guarantee and Interference Mitigation (HGIM) cell-level resource allocation scheme is proposed. Compared with the existing cell-level resource allocation and interference coordination schemes, HGIM differentiates the handover users and grants them with higher priorities in the resource allocation process; HGIM adopts a unified cell division model, which achieves better tradeoff between the algorithm complexity and the effectiveness of the ICI mitigation; Meanwhile, HGIM defines the SPL to optimize the sub-carrier allocation. The performance of HGIM is investigated via a system level simulation and compared with the conventional SFR scheme. Results show that the proposed scheme achieves better handover performance and greater ICI mitigation at the same time. Future work will investigate the HGIM scheme’s performance under more complicated system scenarios, which includes mixed traffic with different QoS requirements, the non-uniform load distribution, and the existence of the handover measurements errors etc.


6

REFERENCES

Inter-Cell Interference (ICI)

5

[1]

4

3

2

1

0 0.1

0.15


0.35

Fig.9 ICI comparison 7000 HGIM Scheme SFR Scheme

System Throughput [kbps]

6000

5000

4000

3000

2000

1000

0 0.05

0.1

0.15 0.2 0.25 0.3 Poisson Arrival Rate(lamda) User/S

0.35

Fig.10 System throughput comparison

Fig.9 and Fig.10 shows that compared with the SFR, HGIM is more effective in term of ICI mitigation because under the

T.S.Rappaport etc, Wireless communications:Past events and a future perspective, IEEE Commun. Mag., vol.40, no.5, pp.148-161, May 2002 [2] 3GPP TSG RAN WG1 Meeting #41 R1-050507 Athens, Greece, Huawei, 9-13 May, 2005 [3] TSG-RAN WG1 Meeting #42 R1-050738 London, UK, Siemens, 29 August- 2 September, 2005 [4] 3GPP TSG RAN WG1, R1-051059, San Diego, USA, Texas Instruments, 10 October - 14 October, 2005 [5] 3GPP TSG RAN WG1 Meeting #43, R1-051341, Seoul, Korea, Samsung, 7-11 November, 2005 [6] Fangming Xu, Xiaofeng Tao, Xiaodong Xu, Soft Fractional Frequency Reuse, ZTE Communications, pp.17-19,April 2007 [7] Zhongnian Li, Yafeng Wang, A Hybrid Inter-cell Interference Mitigation Scheme for OFDMA System, ICCS08, pp. 656-660 [8] Mohmmad, Anas etc, Performance Evaluation of Received Signal Strength Based Hard handover for UTRAN LTE, VTC2007-Spring, pp.1046-1050,22-25 April 2007 [9] Sik Choi, Gyung-Ho Hwang etc, Fast Handover Scheme for Real-Time Downlink Services in IEEE 802.16e BWA System, IEEE VTC 2005spring,vol. 3, pp. 2028-2032, June 2005 [10] Antti Tolli, Marian Codreanu and Markku Juntti, Soft Handover in Adaptive MIMO-OFDM Cellular System with Cooperative Processing, IEEE PIMRC 2006, pp.1-5, 11-14 Sept. 2006 [11] Hyungkeuk Lee , Hyukmin Son , Sanghoon Lee, Semisoft Handover Gain Analysis Over OFDM-Based Broadband Systems, IEEE Trans on Vehicular Tech, vol. 58, no. 3, pp. 1443-1453, Mar 2009 [12] Changchuan Yin, Tao Luo etc, Multi-carrier wireless communication technology. BUPT Press, July 2004